Research map / Detection and segmentation / CNN features
MP-Former: Mask-Piloted Transformer for Image Segmentation
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Add mask-piloted training | Mask2Former · Cityscapes val, R50 instance segmentation | AP: 37.4 -> 39.7 | Helped |
| Add mask-piloted training | Mask2Former · Cityscapes val, R50 semantic segmentation | mIoU: 79.4 -> 81.0 | Helped |
| Add mask-piloted training | Mask2Former · COCO val2017 instance segmentation, R50 | AP: 43.7 -> 44.8 | Helped |
| Add MP training on the first decoder layer (ablation) | Mask2Former, R50 · COCO instance segmentation, 12 epochs | AP: 38.7 -> 39.4 | Helped |
| Add label noise to MP training (ablation) | Mask2Former, R50 · COCO instance segmentation, 12 epochs | AP: 39.4 -> 39.6 | Helped |
| Add label noise to MP training (ablation) | Mask2Former, R50 · COCO instance segmentation, 50 epochs | AP: 44.2 -> 44.2 | No clear effect |
| Add point noise to masks (ablation) | MP training, R50 · COCO instance segmentation, 12 epochs | AP: 39.6 -> 39.9 | Helped |
| Replace point noise with shifting noise (ablation) | MP training, R50 · COCO instance segmentation, 12 epochs | AP: 39.6 -> 39.6 | No clear effect |
| Replace point noise with scaling noise (ablation) | MP training, R50 · COCO instance segmentation, 12 epochs | AP: 39.6 -> 39.4 | Hurt |
| Apply MP training to all decoder layers rather than the first layer (ablation) | MP training, R50 · COCO instance segmentation, 12 epochs | AP: 39.4 -> 40.2 | Helped |
| Use the largest feature map at every decoder layer (ablation) | MaskDN-Former, R50 · COCO instance segmentation, 12 epochs | AP: 40.2 -> 40.0 | Hurt |
| Use matching instead of hard assignment in the MP part (ablation) | MaskDN-Former, R50 · COCO instance segmentation, 12 epochs | AP: 39.6 -> 39.4 | Hurt |
Closest papers by meaning
- Masked-Attention Mask Transformer for Universal Image Segmentation (CVPR 2022 · similarity 0.92)
- Per-Pixel Classification is Not All You Need for Semantic Segmentation (NeurIPS 2021 · similarity 0.89)
- Learning Content-Enhanced Mask Transformer for Domain Generalized Urban-Scene Segmentation (AAAI 2024 · similarity 0.89)
- PEM: Prototype-based Efficient MaskFormer for Image Segmentation (CVPR 2024 · similarity 0.88)
- Rethinking Decoders for Transformer-based Semantic Segmentation: A Compression Perspective (NeurIPS 2024 · similarity 0.88)
- MaskDiff: Modeling Mask Distribution with Diffusion Probabilistic Model for Few-Shot Instance Segmentation (AAAI 2024 · similarity 0.86)
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.