AI Paper Atlas

Research map / Detection and segmentation / CNN features

Deep Regionlets: Blended Representation and Deep Learning for Generic Object Detection

Hongyu Xu, Xutao Lv, Xiaoyu Wang, Zhou Ren, Navaneeth Bodla, Rama Chellappa

TPAMI 2019 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 34 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 the region selection network (multiple regions) instead of one global region (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 30.27 -> 81.3Improved
select non-rectangular regions instead of offset-only rectangular regions (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 78.5 -> 81.3Improved
add gating-based soft regionlet selection (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 81.3 -> 82.0Improved
increase selected regions from 4 to 9 at 4x4 density (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 79.9 -> 80.9Improved
increase selected regions from 9 to 16 at 4x4 density (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 80.9 -> 82.0Improved
increase regionlet density from 5x5 to 6x6 with 16 regions (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 81.6 -> 80.8Hurt
increase regionlet density from 2x2 to 4x4 with 16 regions (ablation)Deep Regionlets · PASCAL VOC2007 test (train 07+12, ablation)mAP@0.5 (%): 80.0 -> 82.0Improved
replace traditional regionlets with end-to-end deep regionlets (AlexNet)Regionlet / DNP · PASCAL VOC2007 test (train 07)mAP@0.5 (%): 41.7 -> 63.2Improved
use deeper backbone ResNet-101 instead of VGG16Ours-A · PASCAL VOC2007 test (train 07)mAP@0.5 (%): 73.0 -> 75.3Improved
apply soft-NMS at test timeOurs-A · PASCAL VOC2007 test (train 07+12, ResNet-101)mAP@0.5 (%): 82.2 -> 83.1Improved
use projective instead of affine transformation in RSNDeep Regionlets · MS COCO 2017 test-devmmAP 0.5:0.95: 39.3 -> 39.9Improved
add more training data (VOC07+12 instead of 07)Ours-A (VGG16) · PASCAL VOC2007 testmAP@0.5 (%): 73.0 -> 79.2Improved

Earlier methods this paper compares with

Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 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