Apppublic

Chirana/Diffusion_model

sourceHugging Faceupdated 8mo agoView on Hugging Face

0likes

App README

LuminaDiff: Low-Light Image Enhancement

LuminaDiff is a diffusion-based model for enhancing low-light images. It uses a custom U-Net architecture (PureDiffusionUNet) and a DiffusionEngine to recover details and correct lighting.

Project Structure

—src/: Core source code (model, config, diffusion engine).
—weights/: Pre-trained model weights.
—infer.py: Main inference script (Guide & Stitch strategy).
—train.py: Training script.

Setup

1.Install dependencies:

bash

   pip install -r requirements.txt

1.Ensure weights are in weights/latest.pth.

Usage

Inference

To run the advanced inference (Guide & Stitch):

bash

python infer.py

This will process images in the test/ folder and save results to results_advanced/.

Docker

Build and run the container:

bash

docker build -t lumina-diff .
docker run --rm -v $(pwd)/test:/app/test -v $(pwd)/results_advanced:/app/results_advanced lumina-diff