Research map / Detection and segmentation / Weakly supervised detection and segmentation
Improving Few-Shot Part Segmentation Using Coarse Supervision
Oindrila Saha, Zezhou Cheng, Subhransu Maji
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 latent-variable EM with keypoint and mask coarse supervision (vs plain fine-tuning) | ImageNet-init ResNet-34 · CUB Part Test | mIoU: 45.37 -> 48.05 | Improved |
| add EM coarse supervision (vs plain fine-tuning) | keypoint-init ResNet-34 · CUB Part Test | mIoU: 41.12 -> 49.25 | Improved |
| add EM coarse supervision (vs plain fine-tuning) | random-init ResNet-34 · CUB Part Test | mIoU: 29.88 -> 37.98 | Improved |
| replace handcrafted PointSup losses with learned dependencies (EM) | keypoint-init ResNet-34 · CUB Part Test | mIoU: 46.45 -> 49.25 | Improved |
| add multi-task keypoint prediction (vs fine-tuning) | ImageNet-init ResNet-34 · CUB Part Test | mIoU: 45.37 -> 41.27 | Hurt |
| add multi-task keypoint prediction (vs fine-tuning) | random-init ResNet-34 · CUB Part Test | mIoU: 29.88 -> 36.96 | Improved |
| add PseudoSup semi-supervised pseudo-labels (vs fine-tuning) | ImageNet-init ResNet-34 · CUB Part Test | mIoU: 45.37 -> 46.01 | Improved |
| add keypoint supervision to mask-only EM (ablation) | EM part segmentation · CUB Part Test | mIoU: 46.30 -> 49.25 | Improved |
| add mask supervision to keypoint-only EM (ablation) | EM part segmentation · CUB Part Test | mIoU: 47.96 -> 49.25 | Improved |
| add EM coarse supervision (vs fine-tuning) | ImageNet-init ResNet-34 · OID Part test | mIoU: 55.30 -> 58.68 | Improved |
| add EM coarse supervision (vs multi-tasking) | ImageNet-init ResNet-34 · OID Part test | mIoU: 55.61 -> 58.68 | Improved |
| add EM coarse supervision (vs fine-tuning) | ImageNet-init ResNet-34 · PASCAL Birds test | mIoU: 34.25 -> 36.31 | Improved |
Closest papers by meaning
- SCOPS: Self-Supervised Co-Part Segmentation (CVPR 2019 · similarity 0.89)
- PartDistillation: Learning Parts From Instance Segmentation (CVPR 2023 · similarity 0.89)
- Learning to Annotate Part Segmentation with Gradient Matching (ICLR 2022 · similarity 0.89)
- From Pose to Part: Weakly-Supervised Pose Evolution for Human Part Segmentation (TPAMI 2022 · similarity 0.87)
- Towards Open-World Segmentation of Parts (CVPR 2023 · similarity 0.87)
- WISH: Weakly Supervised Instance Segmentation using Heterogeneous Labels (CVPR 2025 · 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.
