Table 5 Main properties of state-of-the-art methods
From: A survey of deep learning methods for density estimation and crowd counting
Properties | Multi-column | Single-column | Attention-based | Dilation convolution | Spatial transformer | CRFs/MRF | Perspective information | Pyramid pooling | Pan-density /sub-region |
|---|---|---|---|---|---|---|---|---|---|
Methods | |||||||||
MCNN [1] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | ||||||
CSRNet [12] | \(\checkmark\) | \(\checkmark\) | |||||||
DRSAN [84] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | ||||||
DecideNet [7] | \(\checkmark\) | \(\checkmark\) | |||||||
SCNet [228] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | ||||||
PaDNet [92] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | ||||||
SAAN [8] | \(\checkmark\) | \(\checkmark\) | |||||||
PACNN [98] | \(\checkmark\) | \(\checkmark\) | |||||||
CAN&ECAN [99] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | |||||
SFANet [102] | \(\checkmark\) | \(\checkmark\) | |||||||
W-Net [103] | \(\checkmark\) | \(\checkmark\) | |||||||
DSNet [205] | \(\checkmark\) | \(\checkmark\) | |||||||
L2SM [111] | \(\checkmark\) | ||||||||
DSSINet [220] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | |||||
SPANet [235] | \(\checkmark\) | \(\checkmark\) | |||||||
MBTTBF-SCFB [236] | \(\checkmark\) | \(\checkmark\) | \(\checkmark\) | ||||||
BL [234] | \(\checkmark\) | ||||||||
S-DCNet [237] | \(\checkmark\) | \(\checkmark\) | |||||||
PGCNet [159] | \(\checkmark\) | \(\checkmark\) |