Research map / Detection and segmentation / Open-set and few-shot detection
VizWiz-FewShot: Locating Objects in Images Taken by People with Visual Impairments
Yu-Yun Tseng, Alexander Bell, 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 |
|---|---|---|---|
| increase shots from 1 to 10 | YOLACT FSIS · VizWiz-FewShot-IS mean | mAP: 1.81 -> 6.42 | Improved |
| increase shots from 1 to 10 | DeFRCN · VizWiz-FewShot-OD mean | mAP: 4.03 -> 12.61 | Improved |
| evaluate on VizWiz instead of MS COCO test | DeFRCN · 1-shot cross-dataset | mAP: 6.63 -> 1.32 | Hurt |
| evaluate on VizWiz instead of MS COCO test | DeFRCN · 10-shot cross-dataset | mAP50: 29.15 -> 6.53 | Hurt |
| objects contain holes | DeFRCN · VizWiz-FewShot-OD 10-shot | mAP: 12.49 -> 5.37 | Hurt |
| objects contain text | DeFRCN · VizWiz-FewShot-OD 10-shot | mAP: 15.96 -> 17.45 | Improved |
| high instead of medium image quality | YOLACT FSIS · VizWiz-FewShot-IS 1-shot | mAP: 1.24 -> 2.11 | Improved |
| high instead of medium image quality | DeFRCN · VizWiz-FewShot-OD | mAP: 12.82 -> 13.62 | No clear effect |
| small instead of medium objects | YOLACT FSIS · VizWiz-FewShot-IS 10-shot | mAP: 6.53 -> 3.94 | Hurt |
| tuned hyperparameters (lower LR, loss weights, COCO resolution, size filtering) instead of defaults | YOLACT · VizWiz-FewShot-IS | mAP: around 0 -> 1.81 | Improved |
| compare dataset to COCO-20i on 1-shot | YOLACT · VizWiz-IS vs COCO-20i | mAP50: 17.1 -> 2.48 | Hurt |
| compare dataset to COCO-20i on 5-shot | YOLACT · VizWiz-IS vs COCO-20i | mAP50: 18.9 -> 5.17 | Hurt |
Closest papers by meaning
- SILCO: Show a Few Images, Localize the Common Object (ICCV 2019 · similarity 0.86)
- Meta-Learning to Detect Rare Objects (ICCV 2019 · similarity 0.86)
- FSS-1000: A 1000-Class Dataset for Few-Shot Segmentation (CVPR 2020 · similarity 0.85)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.84)
- Revisiting Few-Shot Object Detection with Vision-Language Models (NeurIPS 2024 · similarity 0.84)
- Task-Disruptive Background Suppression for Few-Shot Segmentation (AAAI 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.
