Neural Radiance Fields (NeRF) have greatly advanced 3D reconstruction with high-quality rendering and realistic detail recovery. Effectively removing occlusions while preserving scene details can further enhance NeRF’s robustness and applicability. However, existing approaches for object and occlusion removal predominantly rely on generative priors, which, despite filling the resulting holes, introduce new artifacts and blurriness. Moreover, existing benchmark datasets for evaluating occlusion removal methods lack realistic complexity and viewpoint variations. To address these issues, we introduce DeclutterSet, a novel dataset featuring diverse scenes with significant occlusions across foreground, midground, and background objects that vary distinctly with viewpoints. We further introduce DeclutterNeRF, an occlusion removal method free from generative priors. DeclutterNeRF introduces joint multiview optimization of learnable camera parameters, occlusion annealing regularization, and employs an explainable stochastic structural similarity loss, ensuring highquality, artifact-free reconstructions from incomplete imagery. Experiments demonstrate that DeclutterNeRF significantly outperforms state-of-the-art methods on our proposed DeclutterSet, establishing a strong baseline for future research. All the code and data will be released.
[A Project on LLM for Autonomous Driving]Under review
