AI Paper Atlas

Research map / Detection and segmentation / Open-set and few-shot detection

AsyFOD: An Asymmetric Adaptation Paradigm for Few-Shot Domain Adaptive Object Detection

Yipeng Gao, Kun-Yu Lin, Junkai Yan, Yaowei Wang, Wei-Shi Zheng

CVPR 2023 · Detection and segmentation › Open-set and few-shot detection · approach: Domain-adaptive object detection · 32 citations (OpenAlex)

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 changedWhereMeasuredEffect
Use AsyFOD instead of AcroFODYOLOv5 X detector · Cityscapes→Foggy CityscapesmAP50: 41.1±0.8 -> 44.3±1.0Improved
Use AsyFOD instead of AcroFODYOLOv5 X detector · Sim10K→CityscapesAP50: 62.5±1.6 -> 65.4±0.9Improved
Use AsyFOD instead of AcroFODYOLOv5 X detector · KITTI→CityscapesAP50: 62.6±2.1 -> 64.1±1.1Improved
Use AsyFOD instead of Pre+FTYOLOv5 X detector · ViPeD→COCOAP50: 43.2±0.8 -> 47.4±0.7Improved
Use AsyFOD instead of Pre+FTYOLOv5 X detector · ViPeD→COCOAP: 21.0±0.5 -> 23.1±0.5Improved
Add stop-gradient on target features during alignment (ablation)AsyFOD alignment · Sim10K→CityscapesAP50: 62.2 -> 63.4Improved
Apply stop-gradient to source features rather than target features (ablation)AsyFOD alignment · Sim10K→CityscapesAP50: 63.4 -> 61.7Hurt
Add target-similar source instances to augment target instances for alignment (ablation)AsyFOD source-instance division · Sim10K→CityscapesAP50: 56.3±1.1 -> 57.4±0.5Improved
Estimate target-similar source instances using K-means rather than L2 distance (ablation)AsyFOD · Sim10K→CityscapesAP50: 64.5±0.7 -> 65.4±0.9Improved
Estimate target-similar source instances using GMM rather than K-means (ablation)AsyFOD · Cityscapes→Foggy CityscapesmAP50: 43.6±0.7 -> 44.3±1.0Improved
Use target-similar source and target instances for classification, while retaining all boxes for regression (ablation)AsyFOD task-oriented supervised training · Sim10K→CityscapesAP50: 62.5±1.6 -> 64.2±1.1Improved
Restrict regression supervision to target-similar source and target instances (ablation)AsyFOD task-oriented supervised training · Sim10K→CityscapesAP50: 64.2±1.1 -> 63.4±0.9Hurt

Later papers that compared with this method

1 later accepted paper reports AsyFOD as a baseline in their results tables; 0 of them report a better number on most of the results they share with it.

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.

Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy