State-of-the-art results, real-time rendering speed, and robust handling of occlusions and appearance changes.
Enables training on images with varying appearance and achieves the same inference speed as 3DGS.
Uses a DINO-based uncertainty predictor to remove occluders and achieve better results.
Matches the real-time rendering speed of 3DGS while achieving state-of-the-art results.
Uses a DINO-based uncertainty predictor to remove occluders and achieve better results.
Handles in-the-wild scenes with varying appearance and occlusions.
Handling occlusions and appearance changes in in-the-wild scenes.
Achieving state-of-the-art results in 3D Gaussian Splatting.
Enabling real-time rendering for in-the-wild scenes.
Removing occluders and achieving better results.
Train the WildGaussians model on a dataset of in-the-wild scenes.
Use the trained model to render in-the-wild scenes in real-time.
Fine-tune the model for specific use cases if necessary.