Research map / 3D vision / 3D generation
MUVA: A New Large-Scale Benchmark for Multi-View Amodal Instance Segmentation in the Shopping Scenario
Zhixuan Li, Weining Ye, Juan Terven, Zachary Bennett, Ying Zheng, Tingting Jiang, Tiejun Huang
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| train with six views instead of one view (same training data) (ablation) | MASFormer · MUVA test | AP: 25.6 -> 39.4 | Improved |
| train with six views instead of one view (ablation) | Baseline (FES only) · MUVA test | AP: 17.8 -> 22.9 | Improved |
| add view-level attention (ablation) | MASFormer baseline · MUVA test | AP: 22.9 -> 28.7 | Improved |
| add instance-level attention (ablation) | MASFormer baseline · MUVA test | AP: 22.9 -> 34.1 | Improved |
| combine view-level and instance-level attention (ablation) | MASFormer · MUVA test | AP: 34.1 -> 39.4 | Improved |
| increase number of input views from 1 to 6 (random order) (ablation) | MASFormer · MUVA test | AP: 25.6 -> 39.4 | Improved |
| randomize order of input views instead of fixed order (ablation) | MASFormer (6 views) · MUVA test | AP: 37.9 -> 39.4 | Improved |
| use stronger backbone (ResNeXt-101 vs ResNet-50) (ablation) | MASFormer (6 views) · MUVA test | AP: 39.4 -> 43.0 | Improved |
| add synthetic MUVA data to real D2SA training data | MASFormer · D2SA val (overlapping categories) | AP: 63.5 -> 68.4 | Improved |
| train on synthetic MUVA only and test on real D2SA | MASFormer · D2SA val (overlapping categories) | AP: 63.5 -> 41.9 | Hurt |
| use Transformer multi-view model instead of BCNet | MASFormer · MUVA test | computational cost: GPU memory 4G -> 13G; FPS 8.3 -> 3.7 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- BCNet — BCNet: Learning Body and Cloth Shape from A Single Image (ECCV 2020) · better on 2 of 3 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Amodal Instance Segmentation with IRAIS Dataset for Sim-to-Real Transfer (ICML 2026 · similarity 0.89)
- Towards Efficient Foundation Model for Zero-shot Amodal Segmentation (CVPR 2025 · similarity 0.86)
- Segment Anything, Even Occluded (CVPR 2025 · similarity 0.86)
- BLADE: Box-Level Supervised Amodal Segmentation through Directed Expansion (AAAI 2024 · similarity 0.85)
- v-CLR: View-Consistent Learning for Open-World Instance Segmentation (CVPR 2025 · similarity 0.85)
- Open-YOLO 3D: Towards Fast and Accurate Open-Vocabulary 3D Instance Segmentation (ICLR 2025 · similarity 0.84)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
