AI Paper Atlas

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

Retrieval Robust to Object Motion Blur

Rong Zou, Marc Pollefeys, Denys Rozumnyi

ECCV 2024 · Detection and segmentation › Open-set and few-shot detection · approach: Object-level representation learning · 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
train on motion-blurred images with blur-aware losses instead of sharp images only (ablation)BRIDGE retrieval model · Synthetic blur-retrieval (no distractors)mAP: 27.57 -> 91.78Improved
add localization loss (ablation)L_con + L_cls · Synthetic (no distractors)mAP: 85.06 -> 90.39Improved
add blur estimation loss (ablation)L_con + L_cls · Synthetic (no distractors)mAP: 85.06 -> 87.17Improved
add blur estimation loss on top of contrastive, classification and localization (ablation)full loss · Synthetic (no distractors)mAP: 90.39 -> 91.78Improved
add contrastive loss to cls + be + loc (ablation)full loss · Synthetic (no distractors)mAP: 91.23 -> 91.78Improved
add classification loss to con + be + loc (ablation)full loss · Synthetic (no distractors)mAP: 87.48 -> 91.78Improved
widen contrastive blur level range from 1 to 5 (ablation)BLISS sample selection · Synthetic (+1M distractors)mAP@100: 82.26 -> 84.09Improved
exclude edge pixels in blur severity (erosion) for blur estimation loss (ablation)blur estimation loss · Synthetic (+1M distractors)mAP@100: 82.83 -> 84.09Improved
use proposed method vs standard retrieval methods on real blurred data without fine-tuningDELG, DOLG, Token baselines · Real-world blur-retrievalmAP: 54.82 -> 62.88Improved
train only on sharp images (Ours-sharp) evaluated with distractors on query BL 1 (ablation)BRIDGE · Synthetic (+1M distractors, query BL 1)mAP@100: 71.93 (query BL 1) vs 32.64 overallMixed

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