Research map / Detection and segmentation / Open-set and few-shot detection
SAQ-SAM: Semantically-Aligned Quantization for Segment Anything Model
Jing Zhang, Zhikai Li, Chengzhi Hu, Xuewen Liu, Qingyi Gu
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 Prompt-Aware Reconstruction (PAR) over PTQ4SAM-L baseline (ablation) | PTQ4SAM-L · COCO instance segmentation (DINO prompts, W4A4), SAM-B | mAP: 14.4 -> 30.2 | Improved |
| add PCC on top of PAR (ablation) | SAQ-SAM · COCO instance segmentation (DINO prompts, W4A4), SAM-B | mAP: 30.2 -> 33.8 | Improved |
| add PCC on top of PAR (ablation) | SAQ-SAM · COCO instance segmentation (YOLOX prompts, W4A4), SAM-B | mAP: 26.2 -> 30.3 | Improved |
| add PCC on top of PAR (ablation) | SAQ-SAM · COCO instance segmentation (DINO prompts, W4A4), SAM-L | mAP: 46.1 -> 46.3 | No clear effect |
| add PCC on top of PAR (ablation) | SAQ-SAM · COCO instance segmentation (Faster R-CNN prompts, W4A4), SAM-L | mAP: 34.6 -> 34.5 | No clear effect |
| use per-stage PAR instead of per-block QDrop reconstruction (ablation) | SAQ-SAM · COCO instance segmentation (YOLOX prompts, W4A4), SAM-B | mAP: 6.6 -> 26.3 | Improved |
| replace QDrop with PAR at per-stage granularity (ablation) | SAQ-SAM · COCO instance segmentation (YOLOX prompts, W4A4), SAM-B | mAP: 17.2 -> 26.3 | Improved |
| use per-stage rather than per-block PAR (ablation) | SAQ-SAM · COCO instance segmentation (YOLOX prompts, W4A4), SAM-B | mAP: 24.4 -> 26.3 | Improved |
| add PCC to QDrop | QDrop · COCO instance segmentation (YOLOX prompts, W4A4), SAM-B | mAP: 13.3 -> 21.2 | Improved |
| add PCC to RepQ-ViT | RepQ-ViT · COCO instance segmentation (YOLOX prompts, W6A6), SAM-B | mAP: 14.5 -> 33.8 | Improved |
| use PCC instead of PTQ4SAM clipping with RepQ-ViT | RepQ-ViT · COCO instance segmentation (YOLOX prompts, W4A4), SAM-B | mAP: 0.4 -> 1.2 | No clear effect |
| add PCC clipping to PTQ4SAM-S statistic pipeline | PTQ4SAM-S · COCO instance segmentation (Faster R-CNN prompts, W6A6), SAM-B | mAP: 15.4 -> 29.8 | Improved |
Closest papers by meaning
- PTQ4SAM: Post-Training Quantization for Segment Anything (CVPR 2024 · similarity 0.93)
- TinySAM: Pushing the Envelope for Efficient Segment Anything Model (AAAI 2025 · similarity 0.89)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.89)
- Stable Segment Anything Model (ICLR 2025 · similarity 0.89)
- SAM-CP: Marrying SAM with Composable Prompts for Versatile Segmentation (ICLR 2025 · similarity 0.89)
- BLO-SAM: Bi-level Optimization Based Finetuning of the Segment Anything Model for Overfitting-Preventing Semantic Segmentation (ICML 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.
