This repository contains modular PyTorch Lightning implementations of popular deep learning models for image classification. All models inherit from a common BaseClassifier class, making it easy to modify, extend, and use for various tasks.

• Modular Design: All models inherit from BaseClassifier, ensuring consistent training, validation, and testing workflows.
• Easy Configuration: Modify hyperparameters like learning rate, batch size, and optimizer directly in the configuration.
• Checkpointing: Automatically saves the best model during training.
• Early Stopping: Prevents overfitting by stopping training if validation performance plateaus.

1. Install Dependencies:

pip install pytorch_lightning torchvision transformers efficientnet-pytorch

2. Benchmarking Models: Use the train_all_models.py script to train and test all models and check what works the best.:

After completion, you will get report in csv as below. Based on the metric, decide which model is appropriate for your task.

3. Train, Test and Inference demo Use the demo.ipynb to follow the whole workflow with a model architecture.

4. GRAD-CAM Visualization

The grad_cam.ipynb notebook provides a detailed workflow for generating Grad-CAM (Gradient-weighted Class Activation Mapping) heatmaps to visualize which regions of an image are most influential for the model's predictions.

To add a new model:

• Create a new Python file under models directory. (e.g., new_model_classifier.py).
• Inherit from BaseClassifier and implement the model-specific logic.
• If needed, Add the new model to the models_to_test dictionary in train_all_models.py.

This project is licensed under the MIT License. See LICENSE for details.