Research Article

Multiscale Meets Spatial Awareness: An Efficient Attention Guidance Network for Human Parsing

Table 1

Embedding two types of Mutual Attention (MA) into four classical SPP models to evaluate them on the ATR test set.

MethodAvg. F1 scoremIoU

DeepLab-v2 [10]73.2158.50
DeepLab-v2 + MA (parallel)73.3661.01
DeepLab-v2 + MA (sequential)73.9461.43
DeepLab-v3 [28]73.3459.41
DeepLab-v3 + MA (parallel)75.4362.41
DeepLab-v3 + MA (sequential)75.4462.39
PSPNet [11]73.4659.27
PSPNet + MA (parallel)75.2762.15
PSPNet + MA (sequential)75.4162.70
Vortex [12]72.9658.55
Vortex + MA (parallel)75.3262.16
Vortex + MA (sequential)75.4962.97