Research map / Detection and segmentation / Open-set and few-shot detection
Seeing Both Sides: Towards Bidirectional Semantic Alignment for Open-Vocabulary Camouflaged Object Segmentation
Guohui Zhang, Fuming Sun, Yu Zhao, Yuqiu Kong, Jing Sun, Fasheng 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 |
|---|---|---|---|
| add SAM prompt encoder and mask decoder to the CLIP baseline (ablation) | BaCLIP baseline · OVCamo | cIoU: 0.311 -> 0.332 | Improved |
| add the CCE module (multi-frequency multi-scale attention) (ablation) | BaCLIP with SAM · OVCamo | cIoU: 0.332 -> 0.428 | Improved |
| add MREM bidirectional cross-attention (ablation) | BaCLIP with SAM and CCE · OVCamo | cIoU: 0.428 -> 0.488 | Improved |
| use bidirectional instead of text-to-visual only attention in MREM (ablation) | MREM · OVCamo | cIoU: 0.462 -> 0.488 | Improved |
| use bidirectional instead of visual-to-text only attention in MREM (ablation) | MREM · OVCamo | cIoU: 0.432 -> 0.488 | Improved |
| add self-attention before the projection in the adaptive SAM prompt (ablation) | BaCLIP · OVCamo | cIoU: 0.470 -> 0.488 | Improved |
| use ConvNeXt-L instead of ViT-L/14 as CLIP backbone (ablation) | BaCLIP · OVCamo | cIoU: 0.458 -> 0.488 | Improved |
| scale backbone from ConvNeXt-L to ConvNeXt-XXL (ablation) | BaCLIP · OVCamo | cIoU: 0.488 -> 0.530 | Improved |
| cascade 3 MFMSA modules rather than 2, 4 or 5 (ablation) | CCE · OVCamo | cIoU: 0.463 (2), 0.488 (3), 0.470 (4), 0.466 (5) | Mixed |
| combine MFCA and MSDA in MFMSA instead of using one alone (ablation) | CCE · OVCamo | cIoU: 0.437 (MSDA), 0.412 (MFCA) -> 0.488 | Improved |
| fine-tune open-vocabulary segmentation baselines on OVCamo after COCO training | OVSeg / CAT-Seg / FC-CLIP · OVCamo | cIoU: CAT-Seg 0.310 -> 0.088 | Hurt |
| BaCLIP vs OVCoser on hard categories | BaCLIP · OVCamo hard categories (N=15) | cIoU: 0.382 -> 0.445 | Improved |
Closest papers by meaning
- Seeing the Unseen: A Semantic Alignment and Context-Aware Prompt Framework for Open-Vocabulary Camouflaged Object Segmentation (ICCV 2025 · similarity 0.95)
- Training-Free Open-Vocabulary Camouflaged Object Segmentation via Fine-Grained Object Binding and Adaptive Hybrid Prompt (CVPR 2026 · similarity 0.92)
- Iterative Refinement of Semantic and Spatial Representations for Open-Vocabulary Camouflaged Object Segmentation (ECCV 2026 · similarity 0.91)
- MM-CamObj: A Comprehensive Multimodal Dataset for Camouflaged Object Scenarios (AAAI 2025 · similarity 0.87)
- Relax Image-Specific Prompt Requirement in SAM: A Single Generic Prompt for Segmenting Camouflaged Objects (AAAI 2024 · similarity 0.87)
- Distilling Spectral Graph for Object-Context Aware Open-Vocabulary Semantic Segmentation (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.
