Research map / Detection and segmentation / Open-set and few-shot detection
Domain Adaptive Object Detection via Asymmetric Tri-way Faster-RCNN
Zhenwei He, Lei Zhang
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 |
|---|---|---|---|
| add the third and fourth conv blocks to the ancillary net (full 4-block vs 1-block) (ablation) | ATF · Cityscapes to Foggy Cityscapes | mAP: 32.3 -> 38.7 | Improved |
| add the third conv block to the ancillary net (full vs 2-block) (ablation) | ATF · Cityscapes to Foggy Cityscapes | mAP: 35.6 -> 38.7 | Improved |
| keep the ancillary net's target features instead of freezing the backbone and fine-tuning on source labels (ablation) | ATF · Cityscapes to Foggy Cityscapes | mAP: 37.3 -> 38.7 | Improved |
| replace parameter-shared source stream with asymmetric ancillary net | ATF vs MAF · Cityscapes to Foggy Cityscapes | mAP: 34.0 -> 38.7 | Improved |
| use ATF instead of source-only Faster-RCNN | Faster-RCNN · Cityscapes to Foggy Cityscapes | mAP: 20.3 -> 38.7 | Improved |
| use ATF instead of source-only Faster-RCNN | Faster-RCNN · KITTI to Cityscapes | car AP: 30.2 -> 42.1 | Improved |
| use ATF instead of source-only Faster-RCNN | Faster-RCNN · Cityscapes to KITTI | car AP: 53.5 -> 73.5 | Improved |
| use ATF instead of source-only Faster-RCNN | Faster-RCNN · Pascal VOC to Clipart | mAP: 27.8 -> 42.1 | Improved |
| use ATF instead of source-only Faster-RCNN | Faster-RCNN · Pascal VOC to Watercolor | mAP: 44.6 -> 54.9 | Improved |
| use ATF instead of source-only Faster-RCNN | Faster-RCNN · SIM10k to Cityscapes | car AP: 34.6 -> 42.8 | Improved |
| raise domain alignment weight alpha from 0.3 to 0.7 | ATF · Cityscapes to Foggy Cityscapes | mAP: 36.7 -> 38.7 | Improved |
| raise alpha from 0.7 to 1.3 | ATF · Cityscapes to Foggy Cityscapes | mAP: 38.7 -> 38.0 | Hurt |
Closest papers by meaning
- Multi-Adversarial Faster-RCNN for Unrestricted Object Detection (ICCV 2019 · similarity 0.92)
- Domain Adaptive Faster R-CNN for Object Detection in the Wild (CVPR 2018 · similarity 0.91)
- Unsupervised Domain Adaptive 3D Detection With Multi-Level Consistency (ICCV 2021 · similarity 0.91)
- Robust Domain Adaptive Object Detection With Unified Multi-Granularity Alignment (TPAMI 2024 · similarity 0.91)
- Domain-Specific Suppression for Adaptive Object Detection (CVPR 2021 · similarity 0.91)
- iFAN: Image-Instance Full Alignment Networks for Adaptive Object Detection (AAAI 2020 · similarity 0.91)
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.
