Research map / Detection and segmentation / Weakly supervised detection and segmentation
Prior to Segment: Foreground Cues for Weakly Annotated Classes in Partially Supervised Instance Segmentation
David Biertimpel, Sindi Shkodrani, Anil S. Baslamisli, Nóra Baka
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 |
|---|---|---|---|
| add object mask prior (OMP) from box-head CAMs to the Mask R-CNN class-agnostic mask head | Our Mask R-CNN (R-50-FPN) · non-voc -> voc | AP: 26.4 -> 36.5 | Improved |
| add object mask prior (OMP) from box-head CAMs to the Mask R-CNN class-agnostic mask head | Our Mask R-CNN (R-50-FPN) · voc -> non-voc | AP: 18.9 -> 31.9 | Improved |
| add OMP | Our Mask R-CNN (R-101-FPN) · voc -> non-voc | AP: 21.0 -> 33.2 | Improved |
| add OMP | Our Mask R-CNN (R-101-FPN) · non-voc -> voc | AP: 27.7 -> 37.1 | Improved |
| add OMP, evaluated on ambiguous RoIs | Our Mask R-CNN (R-50-FPN) · voc -> non-voc ambiguous instances | AP: 6.5 -> 17.1 | Improved |
| add OMP, evaluated on non-ambiguous RoIs | Our Mask R-CNN (R-50-FPN) · Full COCO non-ambiguous instances | AP: 36.2 -> 38.0 | Improved |
| add OMP under full mask supervision on all COCO classes | Our Mask R-CNN · COCO (fully supervised) | AP: +1.6 AP | Improved |
| let mask gradients backpropagate into box head CAMs (OMP vs plain CAM of Mask R-CNN) (ablation) | OPMask · COCO CAM mask AP (train all, test all) | AP: 0.2 -> 8.8 | Improved |
| replace Mask R-CNN with OPMask CAM training on voc classes only, test on non-voc (ablation) | OPMask · COCO CAM mask AP (train voc, test non-voc) | AP: 0.2 -> 3.3 | Improved |
| OPMask versus CPMask with a ResNet-101 backbone | OPMask · voc -> non-voc | AP: 34.0 vs 33.2 | No clear effect |
| OPMask versus ShapeProp with ResNet-50 | OPMask · voc -> non-voc | AP: 30.4 -> 31.9 | Improved |
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.
- ShapeMask — ShapeMask: Learning to Segment Novel Objects by Refining Shape Priors (ICCV 2019) · better on 2 of 2 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
- Learning to Segment Every Thing (CVPR 2018 · similarity 0.93)
- The Surprising Impact of Mask-Head Architecture on Novel Class Segmentation (ICCV 2021 · similarity 0.92)
- BoxInst: High-Performance Instance Segmentation With Box Annotations (CVPR 2021 · similarity 0.92)
- Pointly-Supervised Instance Segmentation (CVPR 2022 · similarity 0.91)
- SIM: Semantic-Aware Instance Mask Generation for Box-Supervised Instance Segmentation (CVPR 2023 · similarity 0.91)
- Extreme Point Supervised Instance Segmentation (CVPR 2024 · similarity 0.89)
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.
