Research map / Detection and segmentation / Open-set and few-shot detection
Refer to Any Segmentation Mask Group With Vision-Language Prompts
Shengcao Cao, Zijun Wei, Jason Kuen, Kangning Liu, Lingzhi Zhang, Jiuxiang Gu, HyunJoon Jung, Liang-Yan Gui, Yu-Xiong Wang
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 |
|---|---|---|---|
| use non-autoregressive decoding instead of autoregressive (ablation) | RAS · ORES MaskGroups-HQ | cIoU: 45.34 -> 53.75 | Improved |
| use non-autoregressive decoding instead of autoregressive (ablation) | RAS · ORES MaskGroups-HQ | latency: 2.13 -> 0.56 | Improved |
| ensemble four visual encoders instead of CLIP only (ablation) | RAS · ORES MaskGroups-HQ | overall cIoU: 52.44 -> 53.75 | Improved |
| ensemble four visual encoders instead of DINOv2 only (ablation) | RAS · ORES MaskGroups-HQ | overall cIoU: 47.71 -> 53.75 | Improved |
| ensemble four visual encoders instead of ConvCLIP only (ablation) | RAS · ORES MaskGroups-HQ | overall cIoU: 53.53 -> 53.75 | Improved |
| use different special tokens for candidate and reference masks instead of none (ablation) | RAS · ORES MaskGroups-HQ | overall cIoU: 50.13 -> 53.75 | Improved |
| use different special tokens instead of the same token (ablation) | RAS · ORES MaskGroups-HQ | overall cIoU: 48.49 -> 53.75 | Improved |
| scale LLM from 7B to 13B (ablation) | RAS · ORES overall cIoU | cIoU: 52.19 -> 53.93 | Improved |
| scale LLM from 7B to 13B (ablation) | RAS · RES average | cIoU: 73.7 -> 75.0 | Improved |
| finetune on MaskGroups-HQ (ORES-FT) | RAS · ORES MaskGroups-HQ overall | cIoU: 53.93 -> 73.13 | Improved |
| finetune on RES data | RAS · RES average | cIoU: 75.0 -> 77.8 | Improved |
| finetune on GRES data | RAS · gRefCOCO average | cIoU: 67.78 -> 71.79 | 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.
- PSALM 1.3B — PSALM: Pixelwise Segmentation with Large Multi-modal Model (ECCV 2024) · better on 2 of 3 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
- Mask Grounding for Referring Image Segmentation (CVPR 2024 · similarity 0.91)
- LQMFormer: Language-aware Query Mask Transformer for Referring Image Segmentation (CVPR 2024 · similarity 0.90)
- GSVA: Generalized Segmentation via Multimodal Large Language Models (CVPR 2024 · similarity 0.90)
- OneRef: Unified One-tower Expression Grounding and Segmentation with Mask Referring Modeling (NeurIPS 2024 · similarity 0.90)
- Unveiling Parts Beyond Objects: Towards Finer-Granularity Referring Expression Segmentation (CVPR 2024 · similarity 0.90)
- SetCon: Towards Open-Ended Referring Segmentation via Set-Level Concept Prediction (NeurIPS 2026 · 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.
