Research map / Detection and segmentation / Open-set and few-shot detection
ProSAM: Enhancing the Robustness of SAM-based Visual Reference Segmentation with Probabilistic Prompts
Xiaoqi Wang, Clint Sebastian, Wenbin He, Liu Ren
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 |
|---|---|---|---|
| replace VRP-SAM prompt encoder with variational t-distribution prompt encoder | VRP-SAM · COCO-20i (mask) | mIoU: 53.28 -> 54.63 | Improved |
| replace VRP-SAM prompt encoder with variational prompt encoder | VRP-SAM · PASCAL-5i (mask) | mIoU: 71.14 -> 72.04 | Improved |
| use Gaussian instead of heavy-tailed t prompt distribution (ablation) | ProSAM · PASCAL-5i (mask) | mIoU: 72.04 -> 71.41 | Hurt |
| increase Monte Carlo samples K from 10 to 15 (ablation) | ProSAM · PASCAL-5i (mask) | mIoU: 72.04 -> 71.42 | Hurt |
| reduce t-distribution degrees of freedom from 5 to 3 (ablation) | ProSAM · PASCAL-5i (mask) | mIoU: 72.04 -> 71.55 | Hurt |
| append two extra linear layers to VRP-SAM to match parameters (ablation) | VRP-SAM · PASCAL-5i F-0 | mIoU: 74.01 -> 74.04 | No clear effect |
| swap ResNet-50 for DINOv2 ViT-B/14 image encoder (ablation) | ProSAM · COCO-20i (mask) | mIoU: 54.63 -> 59.16 | Improved |
| swap ResNet-50 for VGG-16 image encoder (ablation) | ProSAM · PASCAL-5i (mask) | mIoU: 72.04 -> 69.27 | Hurt |
| use sampled prompts with K masks merged instead of mean prompt at inference (ablation) | ProSAM · PASCAL-5i F-0 | mIoU: 75.26 -> 75.16 | No clear effect |
| add Gaussian noise (std 1.2) to prompt embeddings at inference | VRP-SAM vs ProSAM · PASCAL-5i F-0 | mIoU degradation: VRP-SAM about 40% drop vs ProSAM about 20% drop | Improved |
| use variational prompt encoder under COCO to PASCAL domain shift | VRP-SAM · COCO-20i to PASCAL-5i | mean mIoU: 76.44 -> 77.65 | 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.
- VRP-SAM — VRP-SAM: SAM with Visual Reference Prompt (CVPR 2024) · better on 9 of 9 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
- AoP-SAM: Automation of Prompts for Efficient Segmentation (AAAI 2025 · similarity 0.89)
- Rethinking Interactive Image Segmentation with Low Latency High Quality and Diverse Prompts (CVPR 2024 · similarity 0.89)
- SAM-Aware Graph Prompt Reasoning Network for Cross-Domain Few-Shot Segmentation (AAAI 2025 · similarity 0.88)
- Visual Prompting for Generalized Few-shot Segmentation: A Multi-scale Approach (CVPR 2024 · similarity 0.88)
- Segment Anyword: Mask Prompt Inversion for Open-Set Grounded Segmentation (ICML 2025 · similarity 0.88)
- Stable Segment Anything Model (ICLR 2025 · similarity 0.88)
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.
