Exploring Deep Residual Learning for Image Recognition

This project implements ResNet models from scratch and trains them on CIFAR-10, MS COCO, and ImageNet datasets. The performance of the models is evaluated using classification accuracy, precision, recall, and F1-score.

File Structure

project/
└──resnet
    ├── block_config.py
    ├── __init__.py
    ├── models.py
    └── train.py
└──utils
    ├── datasets.py
    ├── evaluations.py
    ├── urls.json
    └── helpers.py
└── main.py
├── downloader.py
├── README.md
├── .gitignore
├── project_report.pdf
└── requirements.txt

Installation

1. Clone the repository:

   git clone https://github.com/mohtasimhadi/resnet_exploration.git
   cd resnet_exploration

2. Create a virtual environment and activate it:

   python -m venv env
   source env/bin/activate

3. Install the necessary libraries:

   pip install -r requirements.txt

Downloading Datasets

ImageNet

If ImageNet dataset is not downloaded, use the following code to download and extract it.

python downloader.py imageNet

MS COCO

If MS COCO dataset is not downloaded, use the following code to download and extract it.

python downloader.py ms_coco

Notes

• CIFAR10 dataset will be downloaded automatically during the training.
• If download fails due to url issues, the urls can be changed from utils/urls.json

Running the Code

python main.py --dataset <dataset_name> --layers <num_layers> [--epochs <num_epochs>] [--batch_size <batch_size>]

Arguments:

• dataset: Specify the dataset to use for training.

  choices: CIFAR10, MSCOCO, ImageNet

• layers: Number of layers in the ResNet model.

• epochs: Number of epochs to train the model (default: 10).

• batch_size: Batch size for training and validation (default: 16).

Example Commands:

Train ResNet on CIFAR-10 with 50 layers for 20 epochs:

python main.py --dataset CIFAR10 --layers 50 --epochs 20 --batch_size 32

Train ResNet on ImageNet with 101 layers for 50 epochs:

python main.py --dataset ImageNet --layers 101 --epochs 50 --batch_size 64

Notes:

• Adjust --batch_size according to your system's GPU memory capacity.

References

@inproceedings{he2016deep,
  title={Deep Residual Learning for Image Recognition},
  author={He, Kaiming and Zhang, Xiangyu and Ren, Shaoqing and Sun, Jian},
  booktitle={Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition},
  year={2016}
}

@article{deng2009imagenet,
  title={ImageNet: A Large-Scale Hierarchical Image Database},
  author={Deng, Jia and Dong, Wei and Socher, Richard and Li, Li-Jia and Li, Kai and Fei-Fei, Li},
  journal={IEEE Computer Vision and Pattern Recognition},
  year={2009}
}

@techreport{cifar10,
  author = {Krizhevsky, Alex and Hinton, Geoffrey},
  title = {Learning Multiple Layers of Features from Tiny Images},
  institution = {University of Toronto},
  year = {2009},
  type = {Technical Report}
}

@inproceedings{lin2014microsoft,
  author = {Lin, Tsung-Yi and Maire, Michael and Belongie, Serge and Hays, James and Perona, Pietro and Ramanan, Deva and Doll{\'a}r, Piotr and Zitnick, C. Lawrence},
  title = {{Microsoft COCO}: Common Objects in Context},
  booktitle = {European Conference on Computer Vision (ECCV)},
  year = {2014}
}