Download
Interactive Deep Colorization is a software project for colorizing black-and-white (grayscale) images using deep learning, allowing users to add a few hints (e.g. scribbles) and get a plausible, fully colorized output. The idea is to merge automatic colorization (via neural networks) with optional user guidance — so if the automatic model’s guess isn’t quite right, the user can nudge colors via hints to steer the result, achieving more controlled, satisfying outputs. The project includes both the older Caffe-based implementation and a more recent PyTorch backend, giving flexibility depending on user preference and infrastructure. Because it handles image reading, hint interpretation, and color mapping internally, users don’t need to build the colorization pipeline from scratch: they only need to supply grayscale images (and optionally hints), and the software produces a full-color version.
Features
- Deep-learning based automatic colorization of grayscale images
- Interactive mode: user-provided hints/scribbles influence color output
- Supports both Caffe and PyTorch backends for flexibility
- Easy to run: simple image input (plus optional hints) → colorized output
- Useful for colorizing old photos, sketches, or grayscale media with minimal effort
- Open-source (MIT license), so easy to adapt, extend or integrate in other tools
Project Samples
