kritarthranjan/quickdraw-ai repository overview

⁠🖼️ Mini Diffusion Model for Sketch Generation

This repository contains a mini diffusion model for sketch generation, trained on the Google QuickDraw dataset. The model is implemented in PyTorch and generates 32×32 grayscale sketches using a lightweight UNet architecture with residual blocks and self-attention.

⁠✨ Features

• 🧠 Custom UNet-based diffusion model
• 🖼️ Inference script for generating and visualizing sketches
• ⚡ Supports both CUDA and CPU
• ⚙️ Configurable sample generation:
  • Number of sketches
  • Save path
  • Other generation parameters

⁠Run inference with:

python inference.py