Research map / Detection and segmentation / Open-set and few-shot detection
Unconstrained Foreground Object Search
Yinan Zhao, Brian Price, Scott Cohen, Danna Gurari
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 |
|---|---|---|---|
| train the background encoder instead of using frozen pretrained weights (ablation) | UFO Search · CAIS (top-100 retrievals) | mAP (%): 13.79 -> 32.17 | Improved |
| add discriminator-filtered training triplets (ablation) | UFO Search · CAIS (top-100 retrievals) | mAP (%): 28.97 -> 32.17 | Improved |
| train with triplet ranking loss instead of MSE regression (ablation) | No Discriminator · CAIS (top-100 retrievals) | mAP (%): 24.87 -> 28.97 | Improved |
| use the decoupled encoder instead of the discriminator at test time (ablation) | UFO Search · CAIS (top-100 retrievals) | mAP (%): 25.77 -> 32.17 | Improved |
| use the encoder instead of the discriminator for retrieval from 60,000 objects (ablation) | UFO Search · 60,000-object database | retrieval time: 365.6 -> 0.1 seconds | Improved |
| add discriminator-guided sampling heuristics (similar scenes, similar objects) (ablation) | training data generation · object database over 60,000 | random samples needed: 731 -> 32 | Improved |
| add discriminator-filtered training triplets (ablation) | UFO Search · CAIS (all retrievals) | mAP (%): 25.69 -> 29.03 | Improved |
| train the background encoder instead of frozen weights (ablation) | UFO Search · CAIS (all retrievals) | mAP (%): 11.54 -> 29.03 | Improved |
| add discriminator-filtered training triplets (ablation) | UFO Search · MS-COCO user study (P@5) | Precision@5 (%): 38.50 -> 41.83 | Improved |
| add discriminator-filtered training triplets (ablation) | UFO Search · MS-COCO user study (P@25) | Precision@25 (%): 35.57 -> 38.83 | Improved |
| use triplet loss instead of regression (ablation) | No Discriminator · MS-COCO user study (P@25) | Precision@25 (%): 30.40 -> 35.57 | Improved |
| use UFO Search instead of the best constrained-search baseline (ablation) | UFO Search vs CFO-C · CAIS (top-100 retrievals), dog | mAP (%): 57.82 -> 36.51 | Hurt |
Closest papers by meaning
- Interpretable Foreground Object Search As Knowledge Distillation (ECCV 2020 · similarity 0.90)
- Compositing-aware Image Search (ECCV 2018 · similarity 0.88)
- FOCUS: Towards Universal Foreground Segmentation (AAAI 2025 · similarity 0.85)
- UFO: A Unified Approach to Fine-grained Visual Perception via Open-ended Language Interface (NeurIPS 2025 · similarity 0.84)
- Interpreting Object-level Foundation Models via Visual Precision Search (CVPR 2025 · similarity 0.84)
- VL-SAM-V2: Open-World Object Detection with General and Specific Query Fusion (NeurIPS 2025 · similarity 0.83)
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.
