AI Paper Atlas

Research map / Label-efficient and robust learning / Test-time adaptation

CD-Buffer: Complementary Dual-Buffer Framework for Test-Time Adaptation in Adverse Weather Object Detection

Youngjun Song, Hyeongyu Kim, Dosik Hwang

CVPR 2026 · Label-efficient and robust learning › Test-time adaptation · approach: Realistic and stable test-time adaptation · 0 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 subtractive buffer with discrepancy-weighted L_mask to BN-only adaptation (ablation)CD-Buffer · Cityscapes foggy 0.02mAP@50: 8.69 -> 16.81Improved
Add discrepancy-weighted L_mask to subtractive buffer (ablation)CD-Buffer · Cityscapes foggy 0.02mAP@50: 15.69 -> 16.81Improved
Use additive buffer instead of subtractive buffer (ablation)CD-Buffer · Cityscapes foggy 0.02mAP@50: 16.81 -> 18.66Improved
Combine additive and subtractive buffers (ablation)CD-Buffer · Cityscapes foggy 0.02mAP@50: 18.66 -> 18.74Improved
Add layer-wise gradient scaling (ablation)CD-Buffer · Cityscapes foggy 0.02mAP@50: 18.74 -> 19.13Improved
Add L_mask to full model with grad scaling (ablation)CD-Buffer · Cityscapes foggy 0.02mAP@50: 18.97 -> 19.13Improved
Regularization weight lambda_reg too low (0.01) or too high (0.5) (ablation)CD-Buffer · hyperparameter sensitivitymAP@50Hurt
Use additive adapters (Buffer TTA) vs subtractive (Pruning TTA) under severe fogbaselines · KITTI fog 50mmAP@50: Buffer TTA 23.21 vs Pruning TTA 33.97Hurt
Use CD-Buffer instead of ActMAD under continual TTATTA · KITTI fog continual (avg)mAP@50: 48.32 -> 55.66Improved
Use CD-Buffer instead of WHWTTA · KITTI rain 75mmmAP@50: 74.33 -> 73.03Hurt
Use CD-Buffer instead of Pruning TTATTA · Cityscapes -> ACDC nightmAP@50: 6.39 -> 8.92Improved

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