Ensemble Convolution Neural Network for Robust Video Emotion Recognition Using Deep Semantics

Table 10

Accuracy of PG-CNN under different amounts of synthetic occlusions.

Images with different scales	PG-CNN accuracy (%)
Occlusion: S8	96.43
Occlusion: S16	95.15
Occlusion: S24	92.46
Images without occlusion	97.02