Multi-Scale Guided Attention Network for Crowd Counting

Table 8

Comparison of the same form of attention mechanism.

Dataset	ShanghaiTech A		ShanghaiTech B
Evaluation	MAE	MSE	MAE	MSE
MCNN	110.2	173.2	26.4	41.3
MCNN + CBAM	120.6	184.61	29.77	49.66
MCNN + MFG	84.4	136.7	14.3	23.3
MGANet (ours)	66.5	96.7	9.2	13.8