AI Paper Atlas

Research map / Detection and segmentation / CNN features

Toward Scale-Invariance and Position-Sensitive Region Proposal Networks

Hsueh-Fu Lu, Xiaofei Du, Ping-Lin Chang

ECCV 2018 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 9 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 global convolutional network (GCN) head (ablation)Baseline RPN · PASCAL VOC 07testAR@1000: from 0.48 to 0.59Improved
add GCN head (ablation)Baseline RPN · COCO minivalAR@1000: from 0.421 to 0.542Improved
add position-sensitive score maps (ablation)Baseline RPN · PASCAL VOC 07testAR@1000: from 0.480 to 0.605 (26%)Improved
add position-sensitive score maps (ablation)GCN-NS · PASCAL VOC 07testAR@1000: from 0.586 to 0.653 (11%)Improved
add position-sensitive score maps (ablation)Baseline RPN · COCO minivalAR@1000: from 0.421 to 0.448 (6%)Improved
add position-sensitive score maps (ablation)GCN-NS · COCO minivalAR@1000: from 0.532 to 0.607 (14%)Improved
add GCN-S head (ablation)Baseline RPN · PASCAL VOC 07testAR small @1000: from 0.254 to 0.471 (85%)Improved
add GCN-NS head (ablation)Baseline RPN · COCO minivalAR small @1000: from 0.308 to 0.422 (37%)Improved
replace separable GCN with large plain kernel (LK-S) at similar parameter budget (ablation)GCN-S · PASCAL VOC 07testAR@1000: not as effective as GCN-S and GCN-NS modelsHurt
use separable-kernel GCN-S instead of Naive extra-CBR3 head (ablation)Naive head · PASCAL VOC 07testAR@1000: similar resultsImproved
add position-sensitivity and GCN together (ablation)Baseline RPN · PASCAL VOC 07testAR@1000: 0.48 -> 0.65 (35%)Improved
add position-sensitivity and GCN together (ablation)Baseline RPN · COCO minivalAR@1000: 0.42 -> 0.61 (44%)Improved

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