NSAI on Sort-of-CLEVR.ipynb

{
 "cells": [
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "# Neuro-Symbolic AI\n",
    "## Sort-of-CLEVR Dataset\n",
    "### Contents\n",
    "- Generate Dataset\n",
    "- Perception Module\n",
    "    - Train\n",
    "    - Inference\n",
    "- Semantic Parser\n",
    "    - Preprocess Data\n",
    "    - Train Seq2Seq Model\n",
    "    - Inference\n",
    "- Program Executor\n",
    "- Plugging Everything Together"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Install Pytorch 1.7 and torchtext 0.8\n",
    "# !pip install torch==1.7.0+cu101 torchvision==0.8.1+cu101 torchtext==0.8.0 -f https://download.pytorch.org/whl/torch_stable.html"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# !git clone https://github.com/nerdimite/neuro-symbolic-ai-soc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": [
    "# cd neuro-symbolic-ai-soc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 1,
   "metadata": {},
   "outputs": [],
   "source": [
    "from data_generator import *\n",
    "import numpy as np\n",
    "import pandas as pd\n",
    "import os\n",
    "import dlib\n",
    "import cv2\n",
    "from skimage.io import imshow\n",
    "from sklearn.metrics.pairwise import cosine_similarity\n",
    "import torch\n",
    "from torch import nn\n",
    "from torchvision.datasets import ImageFolder\n",
    "from torch.utils.data import DataLoader\n",
    "from torchvision import transforms as tfms\n",
    "from PIL import Image"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Dataset Generation"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Breakdown"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 2,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Generate a sample i.e.\n",
    "# a tuple of 1 image, 6 objects' properties, 20 QA vectors (10 relational and 10 non-relational)\n",
    "sample = build_sample()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 3,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Break down the sample\n",
    "img, objects, queries, programs, answers = convert_sample(sample)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 4,
   "metadata": {},
   "outputs": [],
   "source": [
    "img = (img * 255).astype('uint8')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 5,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x19604898488>"
      ]
     },
     "execution_count": 5,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "imshow(cv2.cvtColor(img, cv2.COLOR_BGR2RGB))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 6,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(0, array([ 24, 140]), 'c', (9, 125, 39, 155)),\n",
       " (1, array([30, 79]), 'r', (15, 64, 45, 94)),\n",
       " (2, array([128, 138]), 'c', (113, 123, 143, 153)),\n",
       " (3, array([171, 172]), 'r', (156, 157, 186, 187)),\n",
       " (4, array([128,  23]), 'c', (113, 8, 143, 38)),\n",
       " (5, array([88, 15]), 'c', (73, 0, 103, 30))]"
      ]
     },
     "execution_count": 6,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "objects"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 7,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('What is the closest shape to the orange object?',\n",
       "  'circle',\n",
       "  'filter orange <nxt> relate closest <nxt> query shape'),\n",
       " ('What is the furthest shape from the yellow object?',\n",
       "  'rectangle',\n",
       "  'filter yellow <nxt> relate furthest <nxt> query shape'),\n",
       " ('What is the furthest shape from the yellow object?',\n",
       "  'rectangle',\n",
       "  'filter yellow <nxt> relate furthest <nxt> query shape'),\n",
       " ('What is the furthest shape from the green object?',\n",
       "  'rectangle',\n",
       "  'filter green <nxt> relate furthest <nxt> query shape'),\n",
       " ('What is the furthest shape from the green object?',\n",
       "  'rectangle',\n",
       "  'filter green <nxt> relate furthest <nxt> query shape'),\n",
       " ('What is the furthest shape from the gray object?',\n",
       "  'circle',\n",
       "  'filter gray <nxt> relate furthest <nxt> query shape'),\n",
       " ('How many objects of the same shape as the red object are there?',\n",
       "  '4',\n",
       "  'filter red <nxt> query shape <nxt> filter <nxt> count'),\n",
       " ('What is the closest shape to the red object?',\n",
       "  'rectangle',\n",
       "  'filter red <nxt> relate closest <nxt> query shape'),\n",
       " ('What is the closest shape to the green object?',\n",
       "  'circle',\n",
       "  'filter green <nxt> relate closest <nxt> query shape'),\n",
       " ('How many objects of the same shape as the blue object are there?',\n",
       "  '4',\n",
       "  'filter blue <nxt> query shape <nxt> filter <nxt> count'),\n",
       " ('What is the shape of the yellow object?',\n",
       "  'circle',\n",
       "  'filter yellow <nxt> query shape'),\n",
       " ('Is there a red object on the top?',\n",
       "  'no',\n",
       "  'filter red <nxt> query position <nxt> isTop'),\n",
       " ('Is there a orange object on the left?',\n",
       "  'no',\n",
       "  'filter orange <nxt> query position <nxt> isLeft'),\n",
       " ('What is the shape of the red object?',\n",
       "  'circle',\n",
       "  'filter red <nxt> query shape'),\n",
       " ('Is there a yellow object on the left?',\n",
       "  'yes',\n",
       "  'filter yellow <nxt> query position <nxt> isLeft'),\n",
       " ('Is there a red object on the left?',\n",
       "  'yes',\n",
       "  'filter red <nxt> query position <nxt> isLeft'),\n",
       " ('Is there a green object on the top?',\n",
       "  'yes',\n",
       "  'filter green <nxt> query position <nxt> isTop'),\n",
       " ('Is there a blue object on the left?',\n",
       "  'no',\n",
       "  'filter blue <nxt> query position <nxt> isLeft'),\n",
       " ('Is there a orange object on the left?',\n",
       "  'no',\n",
       "  'filter orange <nxt> query position <nxt> isLeft'),\n",
       " ('Is there a yellow object on the left?',\n",
       "  'yes',\n",
       "  'filter yellow <nxt> query position <nxt> isLeft')]"
      ]
     },
     "execution_count": 7,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(zip(queries, answers, programs))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Generate the full dataset"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [
    {
     "name": "stderr",
     "output_type": "stream",
     "text": [
      "100%|████████████████████████████████████████████████████████████████████████████████| 100/100 [00:08<00:00, 12.24it/s]\n",
      "100%|██████████████████████████████████████████████████████████████████████████████████| 10/10 [00:00<00:00, 12.34it/s]\n"
     ]
    }
   ],
   "source": [
    "# Generate the dataset\n",
    "n_train_imgs = 100\n",
    "n_test_imgs = 10\n",
    "\n",
    "build_dataset(n_train_imgs, data_dir='data/train', prefix='train')\n",
    "build_dataset(n_test_imgs, data_dir='data/test', prefix='test')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Perception Module"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Train the Object Detector"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 11,
   "metadata": {},
   "outputs": [],
   "source": [
    "from perception import *"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 12,
   "metadata": {},
   "outputs": [],
   "source": [
    "def load_image_data(csv_file, img_dir):\n",
    "    '''Load the image detection data for training the object detector'''\n",
    "    # Annotations\n",
    "    annot = pd.read_csv(csv_file)\n",
    "    \n",
    "    data = {}\n",
    "    for i in range(len(os.listdir(img_dir))):\n",
    "        # Get all rows belonging to the same scene\n",
    "        scene = annot[annot['filename'] == f'{i}.jpg']\n",
    "        # Read the image\n",
    "        img = cv2.imread(os.path.join(img_dir, f'{i}.jpg'))\n",
    "        \n",
    "        # Add bounding box information for dlib\n",
    "        bboxes = []\n",
    "        for row in scene.values:\n",
    "            x1, y1, x2, y2 = row[-4:]\n",
    "            dlib_box = dlib.rectangle(left=x1 , top=y1, right=x2, bottom=y2)\n",
    "            bboxes.append(dlib_box)\n",
    "        \n",
    "        data[i] = (img, bboxes)\n",
    "    \n",
    "    return data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 13,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the Image Detection Data\n",
    "det_data = load_image_data('data/train/train_img_det.csv', 'data/train/images')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 19,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Training Results: precision: 1, recall: 1, average precision: 1\n",
      "Saved the model to detector.svm\n",
      "Wall time: 7.45 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# Train the Model\n",
    "train_detector(det_data, 'detector.svm')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Shape Classifier"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 20,
   "metadata": {},
   "outputs": [],
   "source": [
    "class Binarize():\n",
    "    '''PyTorch Transforms Object'''\n",
    "    def __init__(self):\n",
    "        '''Converts Grayscale to Binary (except white every other color is zeroed)'''\n",
    "        pass\n",
    "    \n",
    "    def __call__(self, img_tensor):\n",
    "        '''\n",
    "        Args:\n",
    "            img_tensor (tensor): 0-1 scaled tensor with 1 channel\n",
    "        Returns:\n",
    "            tensor\n",
    "        '''\n",
    "        return (img_tensor > 0.95).float()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 21,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the images\n",
    "shapes_data = ImageFolder('data/shapes_data/', transform=tfms.Compose([tfms.Grayscale(), \n",
    "                                                                       tfms.Resize((40, 40)), \n",
    "                                                                       tfms.ToTensor(),\n",
    "                                                                       Binarize()]))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 22,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "{'circle': 0, 'rectangle': 1}"
      ]
     },
     "execution_count": 22,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "shapes_data.class_to_idx"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 23,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create the dataloader\n",
    "shapes_loader = DataLoader(shapes_data, batch_size=20, shuffle=True)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 24,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Model\n",
    "shape_classifier = ShapeClassifier()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 25,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7a20e3fb463d493d8cec997687fd29c2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 1', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1: Mean Loss = 0.6956230501333872\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e29e92479f0c460988e5ba3370301c85",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 2', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2: Mean Loss = 0.5025369574626287\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7b693d54e7614c789a1bef08edd089b4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 3', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3: Mean Loss = 0.35842590034008026\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7ea794445e18462580452e4a3f1f34d6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 4', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 4: Mean Loss = 0.27837350964546204\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "32a473411b99404a9cdaba3fb28e0e7b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 5', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 5: Mean Loss = 0.20613841464122137\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "501deded58d44019937f9c62bb6afe26",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 6', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 6: Mean Loss = 0.15491939832766852\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "67bc48008f5949db9f6cdc67abd27e48",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 7', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 7: Mean Loss = 0.12177710607647896\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f14778d89d6e45d59e8d7778c0a4cf2f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 8', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 8: Mean Loss = 0.10307637353738149\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9e5063df3bf446dab27e474fd66eb989",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 9', max=6.0, style=ProgressStyle(description_width=…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 9: Mean Loss = 0.08451216295361519\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "513f758a704d4339af50feaccb749727",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 10', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 10: Mean Loss = 0.08357098077734311\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "739c8b63501f4513aa941cbce157e8d0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 11', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 11: Mean Loss = 0.06866028501341741\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "37edffefdd5e4953b4507b93312c8609",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 12', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 12: Mean Loss = 0.0573437474668026\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "174e4e953985487e9d297bc80728ffd6",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 13', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 13: Mean Loss = 0.0476840166375041\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "9f930b2b76c54b6cb0082f655ad26891",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 14', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 14: Mean Loss = 0.04104843859871229\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8d5d751472ca4769aeb4b2c02dff8a02",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 15', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 15: Mean Loss = 0.029985308026274044\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4a930f0b17a342d4af8c7552dd7c808f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 16', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 16: Mean Loss = 0.02983030172375341\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ec672b2bbe7046118ceb97cb38cbb069",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 17', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 17: Mean Loss = 0.06565153691917658\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "57bb4acfa0e64185ae9d1a0cceccf6be",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 18', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 18: Mean Loss = 0.026924270981301863\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1b1d946037be4d23b7267e31759ca8c1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 19', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 19: Mean Loss = 0.025716525502502918\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "e0eaf904d28e424da5c97a66996066b4",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 20', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 20: Mean Loss = 0.01962943933904171\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8a5dc17350e341468540a1d91b227291",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 21', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 21: Mean Loss = 0.021238669597854216\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f3af930d3a4f432baabef143fbb1aef2",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 22', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 22: Mean Loss = 0.013076019939035177\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7179e5d3a7844341aa33ebb33dbb78b0",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 23', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 23: Mean Loss = 0.011313553744306168\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "daed188dade74d8787f500c9d116f10f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 24', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 24: Mean Loss = 0.010856726905331016\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c55b2824c212483c8723c477cff74ed1",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 25', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 25: Mean Loss = 0.009973693328599135\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "6570f0c7f12147c5b8870221d4d81d0d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 26', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 26: Mean Loss = 0.011589121189899743\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b166e7f0dd4c4e9b996b7756990e4101",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 27', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 27: Mean Loss = 0.010315224838753542\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "94c561e08d994a179396907e0bc34865",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 28', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 28: Mean Loss = 0.009830930891136328\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "c6fce8a81ca94614b934170f99967c21",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 29', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 29: Mean Loss = 0.010447487235069275\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "a444f6251c8d4086aeec56f1a671ca1d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 30', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 30: Mean Loss = 0.013001982976372043\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "74b1f48a44c144139c21c3f2cb89c40b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 31', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 31: Mean Loss = 0.010358224467684826\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "beb910f3326b4896812de4c8ee4c7796",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 32', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 32: Mean Loss = 0.014643373743941387\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "903df56078ea49d1b3fa1291ca090a13",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 33', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 33: Mean Loss = 0.009698577768479785\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "cbf557d969e942669e9f1491014abecb",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 34', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 34: Mean Loss = 0.0071504291069383425\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "01b495c6d4d4436b931fa17da32c3b7e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 35', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 35: Mean Loss = 0.01237645762739703\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1c2a64eca92f456594060d7f1713257d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 36', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 36: Mean Loss = 0.00543285661842674\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "221bc8d094714feb88400ecf1c8430a9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 37', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 37: Mean Loss = 0.0049534177912088735\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "4fe6cac5600f4e66b3ce7f3e96889971",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 38', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 38: Mean Loss = 0.004974868769447009\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "827548db07084a659a07bf9a21638730",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 39', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 39: Mean Loss = 0.011749461836492022\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "8e240e55785f4a6eaac63a3d182a7f9d",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 40', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 40: Mean Loss = 0.0037530167028307915\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "40f467f1dffd4348966d46918c6ec05b",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 41', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 41: Mean Loss = 0.0038855269085615873\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f790712d47fc4aa5992d2c151aafcc49",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 42', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 42: Mean Loss = 0.0022858166291067996\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "1e06705951ce404da5e48eea5060e3f9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 43', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 43: Mean Loss = 0.0021262670828339956\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "d9aa48d1a28a4d498d1c253726d10c01",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 44', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 44: Mean Loss = 0.0027932546242179037\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bcb81b41f7344e1895c003f6c44eb650",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 45', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 45: Mean Loss = 0.003645793069154024\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "f3800b254d384f61bb09ccb54882af6e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 46', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 46: Mean Loss = 0.0022263673890847713\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3b172f3ad8674645baf8bd97f37c7042",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 47', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 47: Mean Loss = 0.001351267715411571\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "701ef713da524e46ad75db47e5db111e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 48', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 48: Mean Loss = 0.0012796538721886463\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "509741cef4fd48d59ac9b92330187cb7",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 49', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 49: Mean Loss = 0.0020457161347925044\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "b58250a68e364c60879346f653eff926",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 50', max=6.0, style=ProgressStyle(description_width…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 50: Mean Loss = 0.0010679967720837642\n",
      "\n"
     ]
    }
   ],
   "source": [
    "# Train\n",
    "shape_classifier.train_classifier(shapes_loader, lr=0.0001, epochs=50, filename='classifier.pth')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Perception Pipeline\n",
    "`detector -> attribute (shape and color) extraction -> structural scene representation`"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 26,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the Model\n",
    "perceiver = PerceptionPipe('detector.svm', 'classifier.pth', device='cuda')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Breakdown"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 27,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[(array([[[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         ...,\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [  0, 156, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]]], dtype=uint8),\n",
       "  (171, 172)),\n",
       " (array([[[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         ...,\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]]], dtype=uint8),\n",
       "  (23, 140)),\n",
       " (array([[[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         ...,\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]]], dtype=uint8),\n",
       "  (127, 20)),\n",
       " (array([[[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         ...,\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]]], dtype=uint8),\n",
       "  (127, 136)),\n",
       " (array([[[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [  0, 255,   0],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         ...,\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]]], dtype=uint8),\n",
       "  (31, 80)),\n",
       " (array([[[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         ...,\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]],\n",
       "  \n",
       "         [[255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          ...,\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255],\n",
       "          [255, 255, 255]]], dtype=uint8),\n",
       "  (87, 17))]"
      ]
     },
     "execution_count": 27,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Inference with Detector\n",
    "objects = perceiver.detect(img)\n",
    "objects"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 28,
   "metadata": {},
   "outputs": [],
   "source": [
    "# View the isolated objects (this will not work in colab as its headless)\n",
    "# for obj, _ in objects:\n",
    "#     cv2.imshow('frame', obj)\n",
    "#     cv2.waitKey(0)\n",
    "#     cv2.destroyAllWindows()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 29,
   "metadata": {},
   "outputs": [
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Shape = circle\n",
      "Color = blue\n"
     ]
    }
   ],
   "source": [
    "# Inference with Attribute Extractor (Shape Classifier, Color Extractor)\n",
    "shape, color = perceiver.extract_attributes(objects[3][0])\n",
    "print(f'Shape = {shape}\\nColor = {color}')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Pipeline"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 30,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>shape</th>\n",
       "      <th>color</th>\n",
       "      <th>position</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>rectangle</td>\n",
       "      <td>orange</td>\n",
       "      <td>(171, 172)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>circle</td>\n",
       "      <td>red</td>\n",
       "      <td>(23, 140)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>circle</td>\n",
       "      <td>gray</td>\n",
       "      <td>(127, 20)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>circle</td>\n",
       "      <td>blue</td>\n",
       "      <td>(127, 136)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>rectangle</td>\n",
       "      <td>green</td>\n",
       "      <td>(31, 80)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>circle</td>\n",
       "      <td>yellow</td>\n",
       "      <td>(87, 17)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       shape   color    position\n",
       "0  rectangle  orange  (171, 172)\n",
       "1     circle     red   (23, 140)\n",
       "2     circle    gray   (127, 20)\n",
       "3     circle    blue  (127, 136)\n",
       "4  rectangle   green    (31, 80)\n",
       "5     circle  yellow    (87, 17)"
      ]
     },
     "execution_count": 30,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "scene = perceiver.scene_repr(img)\n",
    "scene"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Semantic Parser"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Preprocess Data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 31,
   "metadata": {},
   "outputs": [],
   "source": [
    "from semantic_parser import *\n",
    "from torchtext.data import BucketIterator"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 32,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create the preprocessor object and preprocess\n",
    "preproc = Preprocessor('data/train/train_q2p.csv')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Get the dataset object\n",
    "train_data = preproc.train_data"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "defaultdict(<bound method Vocab._default_unk_index of <torchtext.vocab.Vocab object at 0x00000196769CABC8>>,\n",
       "            {'<unk>': 0,\n",
       "             '<pad>': 1,\n",
       "             '<sos>': 2,\n",
       "             '<eos>': 3,\n",
       "             '<nxt>': 4,\n",
       "             'filter': 5,\n",
       "             'query': 6,\n",
       "             'shape': 7,\n",
       "             'position': 8,\n",
       "             'relate': 9,\n",
       "             'orange': 10,\n",
       "             'gray': 11,\n",
       "             'count': 12,\n",
       "             'isleft': 13,\n",
       "             'istop': 14,\n",
       "             'closest': 15,\n",
       "             'red': 16,\n",
       "             'yellow': 17,\n",
       "             'furthest': 18,\n",
       "             'blue': 19,\n",
       "             'green': 20})"
      ]
     },
     "execution_count": 34,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Looking at the Vocabulary\n",
    "preproc.prog_f.vocab.stoi"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 35,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "defaultdict(<bound method Vocab._default_unk_index of <torchtext.vocab.Vocab object at 0x0000019662CFEB08>>,\n",
       "            {'<unk>': 0,\n",
       "             '<pad>': 1,\n",
       "             '<sos>': 2,\n",
       "             '<eos>': 3,\n",
       "             'the': 4,\n",
       "             '?': 5,\n",
       "             'object': 6,\n",
       "             'is': 7,\n",
       "             'shape': 8,\n",
       "             'there': 9,\n",
       "             'what': 10,\n",
       "             'a': 11,\n",
       "             'on': 12,\n",
       "             'of': 13,\n",
       "             'orange': 14,\n",
       "             'gray': 15,\n",
       "             'are': 16,\n",
       "             'as': 17,\n",
       "             'how': 18,\n",
       "             'many': 19,\n",
       "             'objects': 20,\n",
       "             'same': 21,\n",
       "             'left': 22,\n",
       "             'top': 23,\n",
       "             'closest': 24,\n",
       "             'to': 25,\n",
       "             'red': 26,\n",
       "             'yellow': 27,\n",
       "             'from': 28,\n",
       "             'furthest': 29,\n",
       "             'blue': 30,\n",
       "             'green': 31})"
      ]
     },
     "execution_count": 35,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Looking at the Vocabulary\n",
    "preproc.que_f.vocab.stoi"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Training"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 34,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Training hyperparameters\n",
    "num_epochs = 3\n",
    "learning_rate = 3e-4\n",
    "batch_size = 8\n",
    "num_steps = len(train_data) / batch_size\n",
    "\n",
    "# Model hyperparameters\n",
    "config = {\n",
    "    'que_vocab_size': len(preproc.que_f.vocab),\n",
    "    'prog_vocab_size': len(preproc.prog_f.vocab),\n",
    "    'embedding_dim': 256,\n",
    "    'num_heads': 8,\n",
    "    'num_encoder_layers': 3,\n",
    "    'num_decoder_layers': 3,\n",
    "    'dropout': 0.10,\n",
    "    'max_len': 20,\n",
    "    'forward_expansion': 4,\n",
    "    'que_pad_idx': preproc.que_f.vocab.stoi[\"<pad>\"]\n",
    "}"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 36,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Save the config as a json file\n",
    "import json\n",
    "with open('config.json', 'w') as f:\n",
    "    json.dump(config, f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 37,
   "metadata": {},
   "outputs": [],
   "source": [
    "device = torch.device(\"cuda\" if torch.cuda.is_available() else \"cpu\")"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 38,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create Training Generator\n",
    "train_loader = BucketIterator.splits((train_data,),\n",
    "                                     batch_size=batch_size,\n",
    "                                     sort_within_batch=True,\n",
    "                                     sort_key=lambda x: len(x.query),\n",
    "                                     device=device)[0]"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 39,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Create Model\n",
    "seq2seq = Seq2Seq(config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 33,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "3cd16a5804b34026ba91e5a6bbd20f8e",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 1', max=250.0, style=ProgressStyle(description_widt…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 1: Mean Loss = 0.22138057363592087\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "7cbfbf5c402745ec95c6d239260cf775",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 2', max=250.0, style=ProgressStyle(description_widt…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 2: Mean Loss = 0.0020437352773733436\n",
      "\n",
      "\n"
     ]
    },
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "ecdf27b016a14ecbabd332e24c2908d9",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, description='Epoch 3', max=250.0, style=ProgressStyle(description_widt…"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "Epoch 3: Mean Loss = 0.0008786344127729535\n",
      "\n",
      "\n",
      "Wall time: 34.6 s\n"
     ]
    }
   ],
   "source": [
    "%%time\n",
    "# Train Model\n",
    "seq2seq.train_model(train_loader, num_epochs, num_steps, filename='semantic_parser.pth')"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "### Inference on Queries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 40,
   "metadata": {},
   "outputs": [],
   "source": [
    "sem_parser = SemanticParser(preproc, config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 41,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['filter green', 'relate furthest', 'query shape']"
      ]
     },
     "execution_count": 41,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "program = sem_parser.predict('What is the furthest shape from the green object?')\n",
    "program"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Program Executor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 42,
   "metadata": {},
   "outputs": [],
   "source": [
    "from program_executor import ProgramExecutor"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 43,
   "metadata": {},
   "outputs": [],
   "source": [
    "executor = ProgramExecutor()"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 44,
   "metadata": {
    "scrolled": true
   },
   "outputs": [
    {
     "data": {
      "text/plain": [
       "['What is the closest shape to the orange object?',\n",
       " 'What is the furthest shape from the yellow object?',\n",
       " 'What is the furthest shape from the yellow object?',\n",
       " 'What is the furthest shape from the green object?',\n",
       " 'What is the furthest shape from the green object?',\n",
       " 'What is the furthest shape from the gray object?',\n",
       " 'How many objects of the same shape as the red object are there?',\n",
       " 'What is the closest shape to the red object?',\n",
       " 'What is the closest shape to the green object?',\n",
       " 'How many objects of the same shape as the blue object are there?',\n",
       " 'What is the shape of the yellow object?',\n",
       " 'Is there a red object on the top?',\n",
       " 'Is there a orange object on the left?',\n",
       " 'What is the shape of the red object?',\n",
       " 'Is there a yellow object on the left?',\n",
       " 'Is there a red object on the left?',\n",
       " 'Is there a green object on the top?',\n",
       " 'Is there a blue object on the left?',\n",
       " 'Is there a orange object on the left?',\n",
       " 'Is there a yellow object on the left?']"
      ]
     },
     "execution_count": 44,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "queries"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 47,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Run the program executor with on a full sample of 20 queries\n",
    "pred_ans = []\n",
    "for que in queries:\n",
    "    program = sem_parser.predict(que)\n",
    "    pred_ans.append(executor(scene, program))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 48,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "[('circle', 'circle'),\n",
       " ('rectangle', 'rectangle'),\n",
       " ('rectangle', 'rectangle'),\n",
       " ('rectangle', 'rectangle'),\n",
       " ('rectangle', 'rectangle'),\n",
       " ('circle', 'circle'),\n",
       " (4, '4'),\n",
       " ('rectangle', 'rectangle'),\n",
       " ('circle', 'circle'),\n",
       " (4, '4'),\n",
       " ('circle', 'circle'),\n",
       " ('no', 'no'),\n",
       " ('no', 'no'),\n",
       " ('circle', 'circle'),\n",
       " ('yes', 'yes'),\n",
       " ('yes', 'yes'),\n",
       " ('yes', 'yes'),\n",
       " ('no', 'no'),\n",
       " ('no', 'no'),\n",
       " ('yes', 'yes')]"
      ]
     },
     "execution_count": 48,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "list(zip(pred_ans, answers))"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "## Plugging Everything Together"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 49,
   "metadata": {},
   "outputs": [],
   "source": [
    "from program_executor import *\n",
    "from perception import *\n",
    "from semantic_parser import *\n",
    "import torch\n",
    "from skimage.io import imshow\n",
    "import cv2"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 52,
   "metadata": {},
   "outputs": [],
   "source": [
    "# Load the config file for transformers\n",
    "import json\n",
    "with open('config.json', 'r') as f:\n",
    "    config = json.load(f)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 53,
   "metadata": {},
   "outputs": [],
   "source": [
    "class NSAIPipeline():\n",
    "    '''End-to-End Pipeline of Neuro-Symbolic AI on Sort-of-CLEVR dataset'''\n",
    "    def __init__(self, \n",
    "                 config,\n",
    "                 detector='models/detector.svm',\n",
    "                 classifier='models/classifier.pth',\n",
    "                 sem_parser='models/semantic_parser.pth',\n",
    "                 train_csv='data/train/train_q2p.csv',\n",
    "                 device=None):\n",
    "        \n",
    "        if device is None:\n",
    "            device = 'cuda' if torch.cuda.is_available() else 'cpu'\n",
    "        self.device = device\n",
    "        \n",
    "        # Perception Module\n",
    "        self.perceiver = PerceptionPipe(detector, classifier, self.device)\n",
    "        \n",
    "        # Semantic Parser\n",
    "        self.preproc = Preprocessor(train_csv)\n",
    "        self.sem_parser = SemanticParser(self.preproc, config, filename=sem_parser, device=self.device)\n",
    "        \n",
    "        # Program Executor\n",
    "        self.executor = ProgramExecutor()\n",
    "        \n",
    "    def predict(self, img, query):\n",
    "        '''\n",
    "        Make Prediction on a single image and question pair\n",
    "        \n",
    "        Args:\n",
    "            img (str/array): pixel values should be in 0-255 range\n",
    "                             of dtype uint8 in BGR color format or\n",
    "                             file path of the image\n",
    "            query (str): question about the image\n",
    "            \n",
    "        Returns:\n",
    "            str: answer of the query\n",
    "        '''\n",
    "        # Load img if it's a path\n",
    "        if type(img) == str:\n",
    "            img = cv2.imread(img)\n",
    "        \n",
    "        # Structured Scene Representation\n",
    "        scene = self.perceiver.scene_repr(img)\n",
    "        # Synthesize Program from Query\n",
    "        program = self.sem_parser.predict(query)\n",
    "        # Execute Program\n",
    "        answer = self.executor(scene, program)\n",
    "        \n",
    "        return answer, program\n",
    "    \n",
    "    def evaluate(self, csv, img_dir, debug=True):\n",
    "        '''\n",
    "        Evaluate the model on a dataset\n",
    "        \n",
    "        Args:\n",
    "            csv (str): path of the csv containing image filename, answer, query and program\n",
    "            img_dir (str): directory containing the images\n",
    "            debug (bool): View the data points which were wrong\n",
    "        Returns:\n",
    "            int: accuracy of the model\n",
    "        '''\n",
    "        data = pd.read_csv(csv).values\n",
    "        \n",
    "        correct = []\n",
    "        for filename, answer, query, program in tqdm(data):\n",
    "            # Load Image\n",
    "            img_path = os.path.join(img_dir, filename)\n",
    "            img = cv2.imread(img_path)\n",
    "            \n",
    "            # Make prediction\n",
    "            pred, pred_prog = self.predict(img, query)\n",
    "            \n",
    "            # Verify answer\n",
    "            if str(pred) == answer:\n",
    "                correct.append(1)\n",
    "            else:\n",
    "                if debug:\n",
    "                    print(filename, answer, pred, query, pred_prog)\n",
    "                correct.append(0)\n",
    "        \n",
    "        acc = (sum(correct) / len(correct)) * 100\n",
    "        \n",
    "        return acc"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 54,
   "metadata": {},
   "outputs": [],
   "source": [
    "nsai = NSAIPipeline(config)"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 55,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "(6, ['filter gray', 'query shape', 'filter', 'count'])"
      ]
     },
     "execution_count": 55,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nsai.predict('data/test/images/1.jpg', 'How many objects of the same shape as the gray object are there?')"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 59,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "application/vnd.jupyter.widget-view+json": {
       "model_id": "bb996bb627b64c19ad9ffc9f3d89981f",
       "version_major": 2,
       "version_minor": 0
      },
      "text/plain": [
       "HBox(children=(FloatProgress(value=0.0, max=200.0), HTML(value='')))"
      ]
     },
     "metadata": {},
     "output_type": "display_data"
    },
    {
     "name": "stdout",
     "output_type": "stream",
     "text": [
      "\n"
     ]
    },
    {
     "data": {
      "text/plain": [
       "99.0"
      ]
     },
     "execution_count": 59,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "# Evaluate on the test and get the accuracy\n",
    "nsai.evaluate('data/test/test_q2p.csv', 'data/test/images/', debug=False)"
   ]
  },
  {
   "cell_type": "markdown",
   "metadata": {},
   "source": [
    "---\n",
    "#### Debugging"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 60,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/plain": [
       "<matplotlib.image.AxesImage at 0x190a4962148>"
      ]
     },
     "execution_count": 60,
     "metadata": {},
     "output_type": "execute_result"
    },
    {
     "data": {
      "image/png": "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\n",
      "text/plain": [
       "<Figure size 432x288 with 1 Axes>"
      ]
     },
     "metadata": {
      "needs_background": "light"
     },
     "output_type": "display_data"
    }
   ],
   "source": [
    "img2 = cv2.imread('data/train/images/2.jpg')\n",
    "imshow(cv2.cvtColor(img2, cv2.COLOR_BGR2RGB))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 61,
   "metadata": {},
   "outputs": [
    {
     "data": {
      "text/html": [
       "<div>\n",
       "<style scoped>\n",
       "    .dataframe tbody tr th:only-of-type {\n",
       "        vertical-align: middle;\n",
       "    }\n",
       "\n",
       "    .dataframe tbody tr th {\n",
       "        vertical-align: top;\n",
       "    }\n",
       "\n",
       "    .dataframe thead th {\n",
       "        text-align: right;\n",
       "    }\n",
       "</style>\n",
       "<table border=\"1\" class=\"dataframe\">\n",
       "  <thead>\n",
       "    <tr style=\"text-align: right;\">\n",
       "      <th></th>\n",
       "      <th>shape</th>\n",
       "      <th>color</th>\n",
       "      <th>position</th>\n",
       "    </tr>\n",
       "  </thead>\n",
       "  <tbody>\n",
       "    <tr>\n",
       "      <th>0</th>\n",
       "      <td>rectangle</td>\n",
       "      <td>orange</td>\n",
       "      <td>(155, 172)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>1</th>\n",
       "      <td>rectangle</td>\n",
       "      <td>red</td>\n",
       "      <td>(31, 108)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>2</th>\n",
       "      <td>circle</td>\n",
       "      <td>gray</td>\n",
       "      <td>(31, 32)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>3</th>\n",
       "      <td>circle</td>\n",
       "      <td>green</td>\n",
       "      <td>(91, 164)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>4</th>\n",
       "      <td>circle</td>\n",
       "      <td>blue</td>\n",
       "      <td>(167, 108)</td>\n",
       "    </tr>\n",
       "    <tr>\n",
       "      <th>5</th>\n",
       "      <td>circle</td>\n",
       "      <td>yellow</td>\n",
       "      <td>(131, 84)</td>\n",
       "    </tr>\n",
       "  </tbody>\n",
       "</table>\n",
       "</div>"
      ],
      "text/plain": [
       "       shape   color    position\n",
       "0  rectangle  orange  (155, 172)\n",
       "1  rectangle     red   (31, 108)\n",
       "2     circle    gray    (31, 32)\n",
       "3     circle   green   (91, 164)\n",
       "4     circle    blue  (167, 108)\n",
       "5     circle  yellow   (131, 84)"
      ]
     },
     "execution_count": 61,
     "metadata": {},
     "output_type": "execute_result"
    }
   ],
   "source": [
    "nsai.executor.scene = nsai.perceiver.scene_repr(img2)\n",
    "nsai.executor.scene"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 78,
   "metadata": {},
   "outputs": [],
   "source": [
    "# program = nsai.sem_parser.predict('How many objects of the same shape as the red object are there? ')\n",
    "# program"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": 80,
   "metadata": {},
   "outputs": [],
   "source": [
    "# nsai.executor.count(nsai.executor.filter_(nsai.executor.query(nsai.executor.filter_('red'), 'shape')))"
   ]
  },
  {
   "cell_type": "code",
   "execution_count": null,
   "metadata": {},
   "outputs": [],
   "source": []
  }
 ],
 "metadata": {
  "kernelspec": {
   "display_name": "Python 3",
   "language": "python",
   "name": "python3"
  },
  "language_info": {
   "codemirror_mode": {
    "name": "ipython",
    "version": 3
   },
   "file_extension": ".py",
   "mimetype": "text/x-python",
   "name": "python",
   "nbconvert_exporter": "python",
   "pygments_lexer": "ipython3",
   "version": "3.7.10"
  }
 },
 "nbformat": 4,
 "nbformat_minor": 4
}