List your company
ResearchFormerly WizWand

Visual Grounding on Flickr30k Entities (test)

86.59AccuracyAMC
25.053241.029157.00572.9809Jun 6, 2016Jun 10, 2017Jun 14, 2018Jun 18, 2019Jun 21, 2020Jun 25, 2021Jun 30, 2022
Updated 8mo ago

Evaluation results

MethodLinks
AMC2022.0686.59——
AMC*training_data_restriction=Visual Genome (VG) box and labels only2022.0686.49——
VMRMDetector=Faster-RCNN (VG)2022.0681.11——
Our approachBackbone=ResNet-1012019.1176.74——
InfoGroundDetector=Faster-RCNN (VG)2022.0676.74——
12-in-1Detector=Faster-RCNN (VG)2022.0676.4——
DDPN2019.1173.3——
Align2GroundDetector=Faster-RCNN (VG)2022.0671——
Our approachBackbone=ResNet-502019.1167.9——
SeqGROUND2019.1161.6——
CITE2019.1159.27——
SS+QRN2019.1155.99——
SPC+PPC2019.1155.49——
IGOP2019.1153.97——
Similarity Network2019.1151.05——
RtP2019.1150.89——
MCBPooling=Multimodal Compact Bilinear, Proposals=Selective Search, Backbone=VGG16, Embedding size=500, d=20482016.0648.69——
MCB2019.1148.69——
Element-wise Product + ConvType=Ablation (Element-wise Product + 2048-D Conv), Proposals=Selective Search, Backbone=VGG16, Embedding size=5002016.0647.86——
Rohrbach et al. (2016)Proposals=Selective Search, Backbone=VGG162016.0647.81——
GroundeR2019.1147.81——
Element-wise ProductType=Ablation (Element-wise Product), Proposals=Selective Search, Backbone=VGG16, Embedding size=5002016.0647.41——
ConcatenationType=Ablation (Simple Concatenation), Proposals=Selective Search, Backbone=VGG16, Embedding size=5002016.0646.5——
Wang et al. (2016)2016.0643.89——
NonlinearSP2019.1143.89——
Plummer et al. (2016)Notes=Footnote mentions 50.89% with box size/color features2016.0643.84——
SMPL2019.1142.08——
Hu et al. (2016b)2016.0627.8——
Plummer et al. (2015)2016.0627.42——
DDPNStage=Two-stage, Visual Encoder=RN1012022.03——73.3
DIGNModule=Two-stage, Backbone=VGG-162022.03—78.73—
FAOAModule=One-stage, Backbone=DarkNet-532022.03—68.71—
FAOAStage=One-stage, Visual Encoder=DN53, Time (ms)=232022.03——68.71
LCMCGModule=Two-stage, Backbone=ResNet-1012022.03—76.74—
QRNetModule=One-stage, Backbone=Swin-S2022.03—81.95—
RefTRStage=One-stage, Visual Encoder=RN101, Time (ms)=402022.03——78.66
ReSC-LargeModule=One-stage, Backbone=DarkNet-532022.03—69.28—
ReSCLStage=One-stage, Visual Encoder=DN53, Time (ms)=362022.03——69.28
SAFFModule=One-stage, Backbone=DarkNet-532022.03—70.71—
SeqTRStage=One-stage, Visual Encoder=DN53, Time (ms)=502022.03——81.23
Similarity netModule=Two-stage, Backbone=ResNet-1012022.03—50.89—
SimilarityNetStage=Two-stage, Visual Encoder=RN101, Time (ms)=1842022.03——60.89
TransVGModule=One-stage, Backbone=ResNet-1012022.03—79.1—
TransVGStage=One-stage, Visual Encoder=RN101, Time (ms)=622022.03——79.1
TransVG (Swin)Module=One-stage, Backbone=Swin-S2022.03—78.18—
ZSGNetStage=One-stage, Visual Encoder=RN502022.03——63.39