This repository contains assignment submissions for the Deep Learning (96943) course, Fall 2024, at Institute of Business Administration (IBA) Karachi.
- DLA3:
- This assignment aims to classify images of cats, dogs, and birds as part of a Kaggle competition. The focus is on experimenting with and optimizing different model architectures, including a custom baseline Convolutional Neural Network (CNN) and pretrained models like VGG-16, ResNet-18/34/50, and EfficientNet-B4.
- The DLA3 folder contains the assignment description PDF, along with the following subfolders:
- Models: Contains .pth files for each model, which store the fine-tuned weights (last layer only) of pretrained models and trained weights from scratch for baseline CNN. The best-performing model's state dictionary was saved.
- Notebook: Contains the Jupyter notebook used for data exploration, model specification, hyperparameter tuning, training, and evaluation.
- Report: Contains a detailed report outlining the methodology, results, and analysis of the project.
- Submissions: Contains the models' predictions in the competition-specified format.
This repo is still in progress...need to upload A0, A1, and A2.