List your company
ResearchFormerly WizWand

Facial Expression Recognition on CK+ Eight Classes (10-fold cross val)

99.3Avg AccuracyPPDN
88.27691.1389496.862Jul 24, 2016Aug 2, 2016Aug 12, 2016Aug 22, 2016Sep 1, 2016Sep 11, 2016Sep 21, 2016
Updated 8mo ago

Evaluation results

MethodLinks
PPDNExperimental Settings=image-based2016.0799.3
PPDN w/o peakExperimental Settings=image-based2016.0799.2
PPDN (standard SGD)Experimental Settings=image-based2016.0799.1
GoogLeNet (baseline)Experimental Settings=image-based2016.0799
DTAGN(Joint)Experimental Settings=sequence-based2016.0797.3
FN2ENTraining Protocol=FaceNet2ExpNet (Fine-Tuned)2016.0996.8
STM-ExpLetExperimental Settings=sequence-based2016.0794.2
3DCNN-DAPExperimental Settings=sequence-based2016.0792.4
AUDNInput Modality=Image-based2016.0992.1
VGG-16 face netTraining Protocol=Fine-Tuned (baseline)2016.0989.9
FN2ENTraining Protocol=Train From Scratch (BN)2016.0988.7