Research map / Detection and segmentation / Open-set and few-shot detection
ORBIT: A Real-World Few-Shot Dataset for Teachable Object Recognition
Daniela Massiceti, Luisa Zintgraf, John Bronskill, Lida Theodorou, Matthew Tobias Harris, Edward Cutrell, Cecily Morrison, Katja Hofmann, Simone Stumpf
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 |
|---|---|---|---|
| meta-train on ORBIT instead of Meta-Dataset (ablation) | ProtoNets · ORBIT CLE-VE | frame accuracy: 58.98 -> 65.16 | Improved |
| meta-train on ORBIT instead of Meta-Dataset (ablation) | CNAPs · ORBIT CLE-VE | frame accuracy: 51.86 -> 66.15 | Improved |
| meta-train on ORBIT instead of Meta-Dataset (ablation) | MAML · ORBIT CLE-VE | frame accuracy: 42.55 -> 70.58 | Improved |
| meta-train on ORBIT instead of Meta-Dataset (ablation) | FineTuner · ORBIT CLE-VE | frame accuracy: 61.01 -> 69.47 | Improved |
| evaluate on clutter target videos instead of clean target videos | ProtoNets · ORBIT | frame accuracy: 65.16 -> 50.34 | Hurt |
| evaluate on clutter target videos instead of clean target videos | FineTuner · ORBIT | frame accuracy: 69.47 -> 53.73 | Hurt |
| evaluate on clutter target videos instead of clean target videos | MAML · ORBIT | frame accuracy: 70.58 -> 51.67 | Hurt |
| sample frames from more context videos per object in train tasks (fixed 96 frames) (ablation) | FineTuner/baselines · ORBIT CLU-VE | frame accuracy | Improved |
| cap train-user objects at 6 or more instead of fewer (ablation) | baselines · ORBIT CLU-VE | frame accuracy | No clear effect |
| meta-train with fewer objects per user than at meta-test (caps of 2, 4, 6, 8) (ablation) | baselines · ORBIT CLU-VE | frame accuracy | Hurt |
| increase number of objects per user in tasks (ablation) | baselines · ORBIT CLU-VE | frame accuracy | Hurt |
Closest papers by meaning
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.87)
- Roboflow100-VL: A Multi-Domain Object Detection Benchmark for Vision-Language Models (NeurIPS 2025 · similarity 0.86)
- Probing the Mid-level Vision Capabilities of Self-Supervised Learning (CVPR 2025 · similarity 0.85)
- UniBench: Visual Reasoning Requires Rethinking Vision-Language Beyond Scaling (NeurIPS 2024 · similarity 0.85)
- General Object Foundation Model for Images and Videos at Scale (CVPR 2024 · similarity 0.85)
- Few-Class Arena: A Benchmark for Efficient Selection of Vision Models and Dataset Difficulty Measurement (ICLR 2025 · similarity 0.85)
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.
