SENetCount: An Optimized Encoder-Decoder Architecture with Squeeze-and-Excitation for Crowd Counting

Table 6

Comparison experiments.

Networks	Architecture	Part_A		Part_B		UCF_QNRF		Mall
		MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE
MCNN	Multicolumn-based	110.2	173.2	26.4	41.4	277.0	426.0	—	—
ic-CNN		68.5	116.2	10.7	16.0	—	—	—	—
SAAN		—	—	16.8	28.4	—	—	1.28	1.68
SPN	Single-column-based	70.0	106.3	9.1	14.6	110.3	184.6	—	—
LA-Batch		65.8	103.6	8.6	13.6	113.0	210.0	1.34	1.60
MANet		65.3	95.5	10.2	16.5	—	—	—	—
AutoScale	Detection-based methods	75.7	150.4	10.4	18.8	124.8	234.7	—	—
LSC-CNN		66.4	117.0	8.1	12.7	120.5	218.3	—	—
SE-ResNeXtCount50+MS-SSIM	Single-column-based	71.8	117.0	8.0	13.2	104.8	182.9	1.19	1.52
SE-ResNeXtCount101+MS-SSIM		71.0	115.0	7.3	12.1	107.7	201.1	1.15	1.49