CelebA Dataset Implementation

This project provides a PyTorch-compatible dataset class for the CelebA dataset, enabling access to images and their associated identities.

Overview

The CelebA class in dataset.py is a custom PyTorch Dataset implementation that loads images from a zip archive (Images.zip) and retrieves image identities from annotation files. It supports training, validation, and testing splits, with optional transforms for images and labels. The display_sample.py script visualizes sample images with their identities.

Files

• dataset.py: Defines the CelebA class for loading and processing the CelebA dataset.
• display_sample.py: Displays a 3x3 grid of sample images with their identities using Matplotlib.
• requirements.txt: Lists required dependencies (torch, torchvision).

Clone Repository

To clone the repository, use:

git lfs clone https://github.com/RocwoDev/CelebA.git

Requirements

Install dependencies with:

pip install -r requirements.txt

Required files (not included):

• Eval/list_eval_partition.txt: Specifies train/val/test splits.
• Anno/identity_CelebA.txt: Maps image names to identities.
• Images.zip: Zip archive containing CelebA images.

Usage

1. Setup Dataset: Ensure the required files are in the correct directory structure:

   .
   ├── Anno
   │   └── identity_CelebA.txt
   ├── Eval
   │   └── list_eval_partition.txt
   ├── Images.zip
   ├── dataset.py
   ├── display_sample.py
   └── requirements.txt
   

2. Run Sample Visualization:

   python display_sample.py

   This displays 9 sample images from the training set with their identities.

3. Use in PyTorch:

   from CelebA import CelebA
   train_dataset = CelebA("train", transform=None, target_transform=None)
   img, identity = train_dataset[0]  # Access image and identity

Notes

• Ensure Images.zip contains images with names matching those in list_eval_partition.txt and identity_CelebA.txt.
• Error handling in __getitem__ raises a ValueError for empty image data but does not handle missing or corrupted images.
• Images are loaded as tensors using torchvision.io.decode_image and may require transforms for model compatibility.

Dependencies

• Python 3.x
• PyTorch
• Torchvision
• Matplotlib (for display_sample.py, it's not required for the dataset class itself)