Color-diffusion
Using diffusion models to color black and white images
0
- Using LAB color space for image coloring
- During model training, only noise is added to the color channel while keeping the brightness channel unchanged
- Using UNet architecture for noise prediction
- Combining grayscale image features with denoised UNet features during training
- Support command-line tools and simple Gradio Web UI for image coloring
- Provided a non Markov forward diffusion process for image coloring
Product Details
Color diffusion is an image coloring project based on a diffusion model, which uses the LAB color space to color black and white images. The main advantage of this project is the ability to utilize existing grayscale information (L channel) and train a model to predict color information (A and B channels). This technology is of great significance in the field of image processing, especially in the restoration of old photos and artistic creation. As an open-source project, the background information shows that Color diffusion was quickly built by the author to satisfy curiosity and experience by training diffusion models from scratch. The project is currently free and there is a lot of room for improvement.
