Skip to content

Latest commit

 

History

History
86 lines (82 loc) · 3.54 KB

README.md

File metadata and controls

86 lines (82 loc) · 3.54 KB
Raw

FlatShapeNet

Dataset for the educational game Ariga. The latest version supports hand-drawn plane shapes and data augmentation: rotation, horizontal-flip, vertical-flip and random erase (dash-line). Older versions can be accessed via the link: github.com/reevald/MathEdu

Dataset

Banner Dataset
Download (122 MB): https://www.kaggle.com/datasets/reevald/geometric-shapes-mathematics
The dataset spans eight classes: circle, kite, parallelogram, rectangle, rhombus, square, trapezoid, triangle. The dataset consists of 20,000 images.

  • Circle: Train(1,500); Val(500); Test(500)
  • Kite: Train(1,500); Val(500); Test(500)
  • Parallelogram: Train(1,500); Val(500); Test(500)
  • Rectangle: Train(1,500); Val(500); Test(500)
  • Rhombus: Train(1,500); Val(500); Test(500)
  • Square: Train(1,500); Val(500); Test(500)
  • Trapezoid: Train(1,500); Val(500); Test(500)
  • Triangle: Train(1,500); Val(500); Test(500)

Generate your own

1. Install requirements

In order to run Processing.py from the command line. For maximum compatability with processing, you'll need to install Java 8 -- specifically, we recommend JRE 8u202 from Oracle.
After installing, ensure that you can run Java by typing the following on the command line:

$ java -version

You should see some output that looks like this:

java version "1.8.0_202"
Java(TM) SE Runtime Environment (build 1.8.0_202-b08)
Java HotSpot(TM) 64-Bit Server VM (build 25.202-b08, mixed mode)

More info about installation Processing: https://py.processing.org/tutorials/command-line/

2. Clone repository

$ git clone https://github.com/reevald/FlatShapeNet.git
$ cd FlatShapeNet/hand_drawn_shape

3. Generate Dataset

Run with default settings

$ java -jar lib/processing-py.jar hand_drawn_shape.pyde

[Optional] Run with custom variable

$ PATH_DEST=D:/ProcessingProject/Resource NUM_TRAIN=100 NUM_VAL=20 NUM_TEST=20 java -jar lib/processing-py.jar hand_drawn_shape.pyde
  • PATH_DEST (default current working directory) option to specify the dataset path to be generated (destination path).
  • NUM_TRAIN (default 1500 images) option to number of train dataset
  • NUM_VAL (default 500 images) option to number of validation dataset
  • NUM_TEST (default 500 images) option to number of test dataset

[Result] Directory structure

📂dataset   
└───📂train
│   └───📂circle
│   └───📂kite
│   └───📂parallelogram
│   └───📂rectangle
│   └───📂rhombus
│   └───📂square
│   └───📂trapezoid
│   └───📂triangle
└───📂val
│   └───📂circle
│   └───📂kite
│   └───📂parallelogram
│   └───📂rectangle
│   └───📂rhombus
│   └───📂square
│   └───📂trapezoid
│   └───📂triangle
└───📂test
    └───📂circle
    └───📂kite
    └───📂parallelogram
    └───📂rectangle
    └───📂rhombus
    └───📂square
    └───📂trapezoid
    └───📂triangle

TODOs

  • Add more custom variable
  • Add more shapes
  • Optimization of dataset generation speed
  • Rewrite in native code (python or java)