Research map / Detection and segmentation / Open-set and few-shot detection
Explicit Semantic–Spatial Alignment for Open-Vocabulary Object Detection
Pengyang Su, Weihong Ren, Shuhuan Han, Qian Dong, Haoran Xu, Xi Ai, Zijian Wang, Zhiyong Wang, Honghai Liu
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 auxiliary DINOv2 spatial stream with explicit alignment (ablation) | CLIP baseline · OV-LVIS | mAP rare: 38.1 -> 39.6 | Improved |
| add Spatial Adapter to the aligned stream (ablation) | ESSA-OVD without FAF · OV-LVIS | mAP rare: 39.6 -> 40.3 | Improved |
| add Frequency-Aware Fusion to adapter model (ablation) | ESSA-OVD · OV-LVIS | mAP rare: 40.3 -> 41.3 | Improved |
| add Spatial Adapter to FAF-only model (ablation) | ESSA-OVD · OV-LVIS | mAP rare: 40.5 -> 41.3 | Improved |
| use Spatial Adapter instead of CBAM adapter (ablation) | ESSA-OVD · OV-LVIS | mAP rare: 41.0 -> 41.3 | Improved |
| use Spatial Adapter instead of LFA adapter (ablation) | ESSA-OVD · OV-LVIS | mAP rare: 40.6 -> 41.3 | Improved |
| use FAF instead of gated fusion (ablation) | ESSA-OVD · OV-LVIS | mAP rare: 40.4 -> 41.3 | Improved |
| use FAF instead of channel fusion (ablation) | ESSA-OVD · OV-LVIS | mAP rare: 40.1 -> 41.3 | Improved |
| use larger DINOv2 ViT-L/14 instead of ViT-B/14 as spatial source (ablation) | ESSA-OVD ViT-L/14 · OV-LVIS | mAP rare: 40.8 -> 41.3 | Improved |
| use larger DINOv2 ViT-L/14 instead of ViT-B/14 as spatial source (ablation) | ESSA-OVD ViT-B/16 · OV-LVIS | mAP rare: 32.2 -> 32.8 | Improved |
| ESSA-OVD vs strongest prior with same backbone | ViT-L/14 · OV-LVIS | mAP rare: 39.3 -> 41.3 | Improved |
| ESSA-OVD vs CCKT-Det++ in cross-dataset transfer | LVIS-trained ViT-L/14 · Objects365 | AP: 15.2 -> 17.1 | 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.
- OV-DQUO — OV-DQUO: Open-Vocabulary DETR with Denoising Text Query Training and Open-World Unknown Objects Supervision (AAAI 2025) · better on 4 of 4 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
- Object-level Semantic and Spatial Distillation for Open Vocabulary Detection (ICML 2026 · similarity 0.93)
- Learning Object-Language Alignments for Open-Vocabulary Object Detection (ICLR 2023 · similarity 0.93)
- Bridging the Gap between Object and Image-level Representations for Open-Vocabulary Detection (NeurIPS 2022 · similarity 0.93)
- Exploring Region-Word Alignment in Built-in Detector for Open-Vocabulary Object Detection (CVPR 2024 · similarity 0.92)
- LLMs Meet VLMs: Boost Open Vocabulary Object Detection with Fine-grained Descriptors (ICLR 2024 · similarity 0.91)
- Cyclic Contrastive Knowledge Transfer for Open-Vocabulary Object Detection (ICLR 2025 · similarity 0.90)
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.
