atlas

Research map / Detection and segmentation / CNN features

R-FCN++: Towards Accurate Region-Based Fully Convolutional Networks for Object Detection

AAAI 2018 · Detection and segmentation › CNN features · approach: Region-based and query-based detectors · 38 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's title only (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 GCM with kernel size 15 to R-FCN (ablation)R-FCN · VOC2007 testmAP: 79.1 -> 80.7Helped
add RCMPooling instead of PSPooling (ablation)R-FCN · VOC2007 testmAP: 79.1 -> 79.6Helped
add RCMPooling on top of GCM (ablation)R-FCN+GCM · COCO minivalAP: 29.5 -> 30.4Helped
enlarge GCM kernel size from 1 to 15 (ablation)GCM · VOC2007 testmAP: 79.1 -> 80.7Helped
replace separable GCM with naive kxk conv (ablation)GCM · VOC2007 testmAP: 80.6 -> 80.6 (k=9)No clear effect
use stack of 3x3 convs instead of large kernel (ablation)GCM · VOC2007 testmAP: 80.6 -> 79.7 (k=9)Hurt
reduce middle channels to 64 (Light-GCM) (ablation)GCM · VOC2007 testmAP: 80.7 -> 80.3Hurt
add RCMPooling across IoU thresholds (ablation)R-FCN · VOC2007 test, IoU 0.9mAP: 12.8 -> 13.6Helped
Light-GCM with ResNet-50 instead of R-FCN ResNet-101R-FCN++ · VOC2007 testtest time (s): 0.093 -> 0.072Helped

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