Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• M²-Occ improves Intersection over Union (IoU) by 4.93% under the safety-critical missing back-view setting compared to baseline methods.result 95%
• M²-Occ maintains performance gains as the number of missing cameras increases, improving IoU by 5.01% with five missing views.result 95%
• The Multi-view Masked Reconstruction (MMR) module recovers missing-view representations directly in feature space by leveraging spatial overlap among neighboring cameras.method 90%
• The Feature Memory Module (FMM) uses a learnable memory bank storing class-level semantic prototypes to refine ambiguous voxel features.method 90%
• Existing camera-based semantic occupancy prediction approaches implicitly assume complete surround-view observations, which rarely holds in real-world deployment.limitation 85%
• M²-Occ achieves performance improvements without compromising full-view performance.result 85%
• The paper introduces a systematic missing-view evaluation protocol on the nuScenes-based SurroundOcc benchmark.method 90%
• The robustness gap of M²-Occ widens as the number of missing cameras increases.result 80%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card