AI Paper Atlas

Research map / Detection and segmentation / Weakly supervised detection and segmentation

PPR-FCN: Weakly Supervised Visual Relation Detection via Parallel Pairwise R-FCN

Hanwang Zhang, Zawlin Kyaw, Jinyang Yu, Shih-Fu Chang

ICCV 2017 · Detection and segmentation › Weakly supervised detection and segmentation · approach: Weakly supervised detection · 136 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 role (subject/object) sensitivity to the score map (ablation)PPR-FCN predicate prediction · VRD predicate predictionR@50: 24.30 -> 42.74Improved
add role sensitivity to the score map (ablation)PPR-FCN predicate prediction · VG predicate predictionR@50: 43.30 -> 61.57Improved
use pairwise RoI pooling with role-sensitive maps instead of position-sensitive map with joint box pooling (ablation)PPR-FCN predicate prediction · VRD predicate predictionR@50: 29.57 -> 42.74Improved
replace pooled scores with fully-connected subnetworks (ablation)PosSeq+Pairwise · VRD predicate predictionR@50: 42.74 -> 47.43Improved
replace pooled scores with fully-connected subnetworks (ablation)PosSeq+Pairwise · VG predicate predictionR@50: 61.57 -> 64.17Improved
add parallel pair-selection branch to classification branch (ablation)PPR-FCN · VRD phrase detectionR@50: 3.56 -> 6.93Improved
add parallel pair-selection branch to classification branch (ablation)PPR-FCN · VG phrase detectionR@50: 0.87 -> 2.41Improved
add parallel pair-selection branch to classification branch (ablation)PPR-FCN · VRD relation detectionR@50: 3.56 -> 5.68Improved
use PPR-FCN spatial-context pairwise pooling instead of VTransE-MILweakly supervised relation detection · VRD relation detectionR@50: 4.28 -> 5.68Improved
use the parallel FCN WSOD module instead of WSDDNweakly supervised object detection · VRD object detectionR@50: 15.08 -> 25.34Improved
use original EdgeBox scores instead of the pre-trained objectness-filtered top 1,000 proposals (ablation)WSOD module · VRD/VG object detectionR@K: drop by about 5 pointsHurt
use fully convolutional pairwise pooling instead of per-region fc subnetworksPPR-FCN · test timems/img: 270 -> 150Improved

Later papers that compared with this method

4 later accepted papers report PPR-FCN as a baseline in their results tables; 4 of them report a better number on most of the results they share with it.

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