Crowd Counting On Shanghaitech A

Metrics

MAE

Results

Performance results of various models on this benchmark

Model Name	MAE	Paper Title	Repository
SANet	67.0	Scale Aggregation Network for Accurate and Efficient Crowd Counting	-
Cascaded-MTL	101.3	CNN-based Cascaded Multi-task Learning of High-level Prior and Density Estimation for Crowd Counting
OrdinalEntropy	65.6	Improving Deep Regression with Ordinal Entropy
SGANet + CL	57.6	Crowd Counting via Segmentation Guided Attention Networks and Curriculum Loss
CAN	62.3	Context-Aware Crowd Counting
GauNet (ResNet-50)	54.8	Rethinking Spatial Invariance of Convolutional Networks for Object Counting
CSRNet-EBC	66.3	CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification
M-SFANet+M-SegNet	57.55	Encoder-Decoder Based Convolutional Neural Networks with Multi-Scale-Aware Modules for Crowd Counting
S-DCNet	58.3	From Open Set to Closed Set: Counting Objects by Spatial Divide-and-Conquer
CSRNet	68.2	CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes
IG-CNN	72.5	Divide and Grow: Capturing Huge Diversity in Crowd Images with Incrementally Growing CNN	-
D-ConvNet	73.5	Crowd Counting With Deep Negative Correlation Learning	-
CP-CNN	73.6	Generating High-Quality Crowd Density Maps using Contextual Pyramid CNNs	-
CLIP-EBC (ViT-B/16)	52.5	CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification
DMCount-EBC	62.3	CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification
LoViTCrowd	54.8	Improving Local Features with Relevant Spatial Information by Vision Transformer for Crowd Counting	-
Liu et al.	73.6	Leveraging Unlabeled Data for Crowd Counting by Learning to Rank
CLIP-EBC (ResNet50)	54.0	CLIP-EBC: CLIP Can Count Accurately through Enhanced Blockwise Classification
PSL-Net	49.9	Crowd Counting and Individual Localization Using Pseudo Square Label	-
Zhang et al.	181.8	Cross-Scene Crowd Counting via Deep Convolutional Neural Networks	-

0 of 34 row(s) selected.