Research map / Detection and segmentation / Open-set and few-shot detection
GAIA: A Transfer Learning System of Object Detection That Fits Your Needs
Xingyuan Bu, Junran Peng, Junjie Yan, Tieniu Tan, Zhaoxiang Zhang
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 |
|---|---|---|---|
| Replace ImageNet pre-training with GAIA pre-training | ResNet50 · COCO minival | AP: 37.08 -> 42.91 | Improved |
| Replace ImageNet pre-training with GAIA pre-training | ResNet101 · COCO minival | AP: 39.41 -> 46.07 | Improved |
| Use 3× ImageNet training instead of 1× ImageNet training | ResNet50 · COCO minival | AP: 37.08 -> 39.78 | Improved |
| Use GAIA pre-training instead of ImageNet pre-training | ResNet50 with DCN and Cascade-Head · COCO minival | AP: 44.9 -> 47.9 | Improved |
| Add task-specific architecture selection | GAIA · COCO minival | AP: 47.9 -> 49.1 | Improved |
| Replace ImageNet pre-training with GAIA pre-training | ResNet50 · Objects365 | mmAP: 21.5 -> 24.0 | Improved |
| Add task-specific architecture selection | GAIA · Objects365 | mmAP: 24.0 -> 26.1 | Improved |
| Replace ImageNet pre-training with GAIA pre-training | ResNet50 · Open Images | mAP: 52.2 -> 59.5 | Improved |
| Add task-specific architecture selection | GAIA · Caltech | MR⁻²: 2.2 -> 1.7 | Improved |
| Replace ImageNet pre-training with GAIA pre-training | ResNet50 · UODB | Average mAP: 64.6 -> 71.9 | Improved |
| Add task-specific architecture selection | GAIA · UODB | Average mAP: 71.9 -> 74.4 | Improved |
| Select random upstream data rather than using no selected data (ablation) | GAIA · KITTI, Comic, Watercolor (10-image setting) | Average mAP: 47.8 -> 45.4 | Hurt |
Closest papers by meaning
- GAIA-Universe: Everything is Super-Netify (TPAMI 2023 · similarity 0.91)
- "A Simple Approach and Benchmark for 21,000-Category Object Detection" (ECCV 2022 · similarity 0.87)
- Learning Rich Features at High-Speed for Single-Shot Object Detection (ICCV 2019 · similarity 0.87)
- End-to-End Object Detection with Transformers (ECCV 2020 · similarity 0.86)
- Deep Learning for Generic Object Detection: A Survey (IJCV 2019 · similarity 0.86)
- Simple Open-Vocabulary Object Detection with Vision Transformers (ECCV 2022 · similarity 0.86)
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.
