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

Table 7

Transfer experiments.

Networks	Part_A → Part_B		Part_B → Part_A		Part_A → UCF_QNRF		UCF_QNRF→Part_A		UCF_QNRF→Part_B
	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE	MAE	RMSE
MCNN	85.2	142.3	221.4	357.8	—	—	—	—	—	—
SPN	23.8	44.2	131.2	219.3	236.3	428.4	87.9	126.3	—	—
LSC-CNN	21.2	33.1	150.2	244.6	198.5	359.1	97.0	154.6	11.6	21.0
SE-ResNeXtCount50+MS-SSIM	15.7	28.5	149.2	231.8	166.2	349.9	79.1	121.7	13.9	19.8
SE-ResNeXtCount101+MS-SSIM	20.0	31.3	147.3	229.5	169.0	362.7	89.2	130.8	14.3	22.3