CIB-EfficientNet

In this repository we implement CIB-EfficientNet (Concise Information in Bottleneck EfficientNet). CIB is method used to increase explainability presented in this Bachelor's Thesis (TBA).

Table of Contents

• CIB-EfficientNet
  • Table of Contents
  • Project Structure
  • Setup
    • Prerequisites
    • Installation
  • Usage
    • Training
    • Testing
    • Notebooks
  • Configuration
  • Data Structure
  • Scripts
  • License
  • Contributing

Project Structure

This repo presents this structure:

model/
	loss.py
	metric.py
	model.py
notebooks/
	attributions.ipynb
	plotter.ipynb
data_loader/
	data_loaders.py
train.py
test.py
config.json

Setup

Prerequisites

• Python 3.9+
• CUDA-compatible GPU (recommended)
• Conda package manager

Installation

1. Clone the repository:

git clone https://github.com/SH02-tech/CIB-EfficientNet.git
cd CIB-EfficientNet

2. Create and activate the conda environment:

conda env create -f environment.yml
conda activate fcrp

3. Verify the installation:

python -c "import torch; print(torch.cuda.is_available())"

Usage

Training

To train the model, use the training script with a configuration file:

python train.py --config config_prostate.json

Available configuration files:

• config_prostate.json - Full prostate dataset configuration
• config_prostate_reduced.json - Reduced prostate dataset configuration
• config_prostate_xmi.json - XMI prostate dataset configuration
• config_prostate_xmi_reduced.json - Reduced XMI prostate dataset configuration

Testing

To evaluate the trained model:

python test.py --config config_prostate.json --resume path/to/checkpoint.pth

Notebooks

The project includes several Jupyter notebooks for analysis and visualization:

• attributions.ipynb - Attribution analysis and visualization
• plottter.ipynb - Plotting utilities and visualizations
• test_results_helper.ipynb - Helper functions for test result analysis
• test_results.ipynb - Test results analysis and visualization

Configuration

The configuration files contain parameters for:

• Model architecture settings
• Training hyperparameters
• Dataset paths and preprocessing options
• Logging and checkpoint settings

Data Structure

The project expects data to be organized according to the splits defined in data_split/:

• train.txt / train_reduced.txt - Training set file lists
• val.txt / val_reduced.txt - Validation set file lists
• test.txt / test_reduced.txt - Test set file lists
• classes.txt - Class labels definition

Scripts

Utility scripts are available in the scripts/ directory:

• train.sh - Training script wrapper
• test.sh - Testing script wrapper
• create_sets_files.sh - Data split creation utility
• tensorboard_last_n.sh - TensorBoard logging utility

License

This project is licensed under the terms specified in the LICENSE file.

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.