AI Paper Atlas

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

Multi-Source Domain Adaptation for Object Detection

Xingxu Yao, Sicheng Zhao, Pengfei Xu, Jufeng Yang

ICCV 2021 · Detection and segmentation › Open-set and few-shot detection · approach: Domain-adaptive object detection · 62 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
add low-level domain discriminator loss (ablation)DMSN source subnets · BDD100k daytime+night -> dawn/duskmAP (%): 28.0 -> 29.9Improved
add high-level domain discriminator loss (ablation)DMSN · BDD100k daytime+night -> dawn/duskmAP (%): 29.9 -> 31.7Improved
add pseudo subnet (ablation)DMSN · BDD100k daytime+night -> dawn/duskmAP (%): 31.7 -> 32.9Improved
add instance-level RPN consistency regularization (ablation)DMSN · BDD100k daytime+night -> dawn/duskmAP (%): 32.9 -> 34.2Improved
add dynamic weighting of source subnets (ablation)DMSN · BDD100k daytime+night -> dawn/duskmAP (%): 34.2 -> 35.0Improved
combine Cityscapes and KITTI sources instead of Cityscapes only (source-only)Faster R-CNN · Cityscapes+KITTI -> BDD100k daytimeAP on car (%): 44.6 -> 43.2Hurt
combine sources instead of Cityscapes-only DASW-ICR-CCR · Cityscapes+KITTI -> BDD100k daytimeAP on car (%): 46.5 -> 43.6Hurt
use DMSN instead of best source-combined DADMSN · BDD100k daytime+night -> dawn/duskmAP (%): 30.9 -> 35.0Improved
use DMSN instead of multi-source MDANDMSN · BDD100k daytime+night -> dawn/duskmAP (%): 27.6 -> 35.0Improved
use DMSN instead of M3SDA on cross-camera adaptationDMSN · Cityscapes+KITTI -> BDD100k daytimeAP on car (%): 44.1 -> 49.2Improved
combine day and night sources in source-only training instead of daytime onlyFaster R-CNN · BDD100k daytime+night -> dawn/duskmAP (%): 30.4 -> 28.9Hurt
apply image-level multi-source alignment (MDAN) instead of single-source DAMDAN · BDD100k daytime+night -> dawn/duskmAP (%): 31.8 -> 27.6Hurt

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