AI Paper Atlas

Research map / Detection and segmentation / CNN features

Bucketed Ranking-based Losses for Efficient Training of Object Detectors

Feyza Yavuz, Baris Can Cam, Adnan Harun Dogan, Kemal Oksuz, Emre Akbas, Sinan Kalkan

ECCV 2024 · Detection and segmentation › CNN features · approach: Box regression and loss design · 3 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
replace RS Loss with bucketed BRS LossCascade R-CNN R-50 · COCO valiteration time (s): 1.28 -> 0.24Improved
replace RS Loss with bucketed BRS LossFaster R-CNN R-50 · COCO valiteration time (s): 0.50 -> 0.17Improved
replace RS Loss with bucketed BRS LossFaster R-CNN R-50 · COCO valAP: 39.4 -> 39.5No clear effect
replace AP Loss with bucketed BAP LossATSS R-50 · COCO valAP: 38.1 -> 38.5Improved
replace AP Loss with bucketed BAP LossPAA R-50 · COCO valAP: 37.3 -> 37.2No clear effect
replace RS Loss with BRS Loss on instance segmentationMask R-CNN · COCO, Cityscapes, LVISiteration time (s): up to ~2.5x fasterImproved
spend saved time on longer training (36 epochs) with BRS vs RS at 12 epochsCascade R-CNN · COCO valAP: 41.1 -> 42.3Improved
spend saved time on longer training (27 epochs) with BRS vs RS at 12 epochsMask R-CNN · COCO valmask AP: 36.3 -> 37.3Improved
bucket negatives in AP LossAP Loss on synthetic logits · synthetic dataloss runtime: up to ~40x speed-upImproved
replace Co-DETR training loss with BRS LossCo-DETR ResNet50 · COCO valAP: 49.3 -> 50.1Improved

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