This repository implements a Denoising Diffusion Probabilistic Model (DDPM) for image generation using TensorFlow and Keras. The model is trained on the Oxford-IIIT Pet dataset and includes a custom-implemented time-aware U-Net model with ResNet blocks and self-attention mechanism.

How to Run

1. Clone the repository and dataset:

   git clone https://github.com/Shreehar01/Image-Generator.git
   cd Image-Generator
   git clone https://github.com/ml4py/dataset-iiit-pet.git

2. Install dependencies:

   pip install -r requirements.txt

3. Start training:

   python main.py

   The script will create a directory named epoch_outputs and save generated images there every 10 epochs.