Research map / Detection and segmentation / Open-set and few-shot detection
Low-shot Object Learning with Mutual Exclusivity Bias
Ngoc Anh Thai, Ahmad Humayun, Stefan Stojanov, Zixuan Huang, Bikram Boote, James Matthew Rehg
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 |
|---|---|---|---|
| contrastive multi-view multi-object finetuning on ABC | DINOv2 ViT B/14 · Toys4k LSME | LSA: 39.24 -> 47.70 | Improved |
| contrastive multi-view multi-object finetuning on ABC | DINOv2 ViT B/14 · Toys4k LSME | SA: 50.88 -> 61.32 | Improved |
| contrastive multi-view multi-object finetuning on ABC | DINOv2 ViT B/14 · ShapeNetCore.v2 LSME | LSA: 32.55 -> 41.03 | Improved |
| contrastive multi-view multi-object finetuning on ABC | DINOv2 ViT B/14 · ShapeNetCore.v2 LSME | SA: 47.68 -> 58.92 | Improved |
| train on Toys4k base-class scenes instead of ABC scenes (same scene count) (ablation) | Ours-DINOv2 · Toys4k LSME | LSA: 47.70 -> 43.62 | Hurt |
| train on Toys4k base-class scenes vs no finetuning (ablation) | DINOv2 ViT B/14 · Toys4k LSME | LSA: 39.24 -> 43.62 | Improved |
| replace FreeSOLO segmenter with FreeSOLO finetuned on ABC (mIoU 0.52 -> 0.81) (ablation) | Ours-DINOv2 B/14-ABC · Toys4k LSME | LSA: 35.50 -> 47.70 | Improved |
| replace SAM segmenter with FreeSOLO-ABC (mIoU 0.72 -> 0.81, a 0.1 gain) (ablation) | Ours-DINOv2 B/14-ABC · Toys4k LSME | SA: 56.92 -> 61.32 | Improved |
| replace CutLer segmenter with SAM (ablation) | Ours-DINOv2 B/14-ABC · Toys4k LSME | LSA: 39.42 -> 42.38 | Improved |
| require mutual exclusivity (support assignment) instead of given shot labels | DINOv2 ViT B/14 · Toys4k | LSA: 57.92 -> 43.21 | Hurt |
| add instance segmentation (full LSME) on top of Categ-MObj-SuppAssign | DINOv2 ViT B/14 · Toys4k | LSA: 43.21 -> 39.24 | Hurt |
| add occlusion from multiple objects (Categ-SObj-PoseVar to Categ-MObj) | DINOv2 ViT B/14 · Toys4k | LSA: 75.18 -> 57.92 | Hurt |
Closest papers by meaning
- Mutual exclusivity as a challenge for deep neural networks (NeurIPS 2020 · similarity 0.86)
- Benchmarking Spurious Bias in Few-Shot Image Classifiers (ECCV 2024 · similarity 0.86)
- Rethinking Few-Shot Adaptation of Vision-Language Models in Two Stages (CVPR 2025 · similarity 0.85)
- Association of Objects May Engender Stereotypes: Mitigating Association-Engendered Stereotypes in Text-to-Image Generation (NeurIPS 2024 · similarity 0.85)
- DiPEx: Dispersing Prompt Expansion for Class-Agnostic Object Detection (NeurIPS 2024 · similarity 0.84)
- AMU-Tuning: Effective Logit Bias for CLIP-based Few-shot Learning (CVPR 2024 · similarity 0.84)
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.
