/ProjectXUBC2023

Primary LanguagePython

Self Expanding CNN (ProjectX UBC submission)

This is the official implementation of the Self Expanding Convolutional Neural Network paper, which is our submission to the ProjectX machine learning research competition organized by UofT AI.

Project Structure

├─models
  ├── __init__.py
  ├── convBlock.py
  ├── dynamicCNN.py
  ├── identityConv.py
  ├── perceptron.py
├─utils
  ├── __init__.py
  ├── train.py
  ├── utils.py
├─ dataloaders.py
├─ main.py
├─ README.md

To Recreate the results:

  • clone the current repository
  • Install the following dependencies in a virtual environment:
    • pytorch
    • wandb
    • tqdm
  • Run the main.py file from the root directory

We benchmarked on the CIFAR 10 dataset. You can easily create your own custom dataset and pass the dataloaders into the train_model function.

HAPPY CODING!!