Research map / Detection and segmentation / Open-set and few-shot detection
Learning To Detect and Segment for Open Vocabulary Object Detection
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Condition the box-regression and mask-segmentation heads on semantic embeddings | RegionCLIP open-vocabulary detector · COCO | detection AP, Novel: 31.3 -> 33.7 (+2.4) | Helped |
| Condition the box-regression and mask-segmentation heads on semantic embeddings | RegionCLIP open-vocabulary detector · COCO | instance segmentation AP, Novel: 27.5 -> 29.7 (+2.2) | Helped |
| Condition detection heads on semantic embeddings | RegionCLIP, RN50-C4 · LVIS | detection APr: 17.0 -> 19.9 (+2.9) | Helped |
| Condition segmentation heads on semantic embeddings | RegionCLIP, RN50-C4 · LVIS | instance segmentation APr: 17.4 -> 20.0 (+2.6) | Helped |
| Condition detection heads on semantic embeddings | ViLD trained on LVIS · COCO cross-dataset evaluation | AP50: 55.6 -> 59.1 (+3.5) | Helped |
| Condition box and mask heads on semantic embeddings | ViLD on LVIS · LVIS | detection APr: 16.6 -> 18.8 (+2.2) | Helped |
| Add CondHead to the box-regression head (ablation) | ViLD* on LVIS · LVIS | detection APr: 16.6 -> 19.6 (+2.7) | Helped |
| Add CondHead to the mask-segmentation head (ablation) | ViLD* on LVIS · LVIS | instance segmentation APr: 16.9 -> 20.0 (+2.2) | Helped |
| Use naive class-wise transfer based on the most semantically similar base category instead of CondHead (ablation) | ViLD · LVIS | box APr: 18.8 -> 17.1 | Hurt |
| Add Shapemask shape priors to CondHead (ablation) | ViLD · LVIS | mask APr: 16.9 -> 20.4 | Helped |
| Add Shapemask shape priors to CondHead (ablation) | RegionCLIP · LVIS | mask APr: 17.4 -> 21.0 | Helped |
| Increase the number of expert heads beyond 40 (ablation) | CondHead · LVIS | detection and segmentation performance | No clear effect |
Closest papers by meaning
- CAKE: Category Aware Knowledge Extraction for Open-Vocabulary Object Detection (AAAI 2025 · similarity 0.90)
- ProxyDet: Synthesizing Proxy Novel Classes via Classwise Mixup for Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.89)
- Boosting Segment Anything Model Towards Open-Vocabulary Learning (AAAI 2025 · similarity 0.89)
- Mask-Adapter: The Devil is in the Masks for Open-Vocabulary Segmentation (CVPR 2025 · similarity 0.89)
- DitHub: A Modular Framework for Incremental Open-Vocabulary Object Detection (NeurIPS 2025 · similarity 0.88)
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (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.