houyuanchen111/awesome-relight

Awesome Relight is not a piece of software you install and run. It is a curated list, a single page of organized links, pointing to research papers, datasets, benchmarks, demos, and tools about relighting. Relighting means changing how light falls on something in an image, video, or 3D scene after it was captured or generated, so a photo taken at noon could be made to look like it was shot at sunset, or a portrait could be lit from a different angle without reshooting it. The list covers this topic from many angles: general surveys of the field, relighting of whole scenes and photos, relighting for video, and relighting focused specifically on portraits, faces, and human bodies. It also tracks work on 3D representations such as NeRF and Gaussian splatting, relighting for driving and robotics footage, security and robustness concerns tied to relighting, and even reinforcement learning approaches. Beyond papers, it collects datasets and benchmarks researchers use to test their methods, workshops and challenges from academic conferences, and actual software, demos, and products people can try. Each entry typically links out to the paper itself, an arXiv preprint, a project website, or a GitHub code repository, so a reader can go from a one line description straight to the source. The project describes its own goal as organizing this work by task and representation rather than just listing things by publication date, so someone new to the topic can find the corner of the field that matches what they are trying to build or study. The repository is released under the CC0-1.0 license, meaning it is placed in the public domain with no restrictions on reuse. It has 14 stars and welcomes contributions through pull requests for missing papers, datasets, code releases, or software links, following a separate contributing guide. There is no programming language associated with the project since it is a reference list rather than a codebase. The full README is longer than what was shown.