Awesome Crowd Counting

Papers

2018

• Crowd Counting with Deep Negative Correlation Learning (CVPR2018) [paper] [code]
• Divide and Grow: Capturing Huge Diversity in Crowd Images with Incrementally Growing CNN (CVPR2018) [paper]
• Structured Inhomogeneous Density Map Learning for Crowd Counting (arXiv) [paper]
• Body Structure Aware Deep Crowd Counting (TIP2018) [paper]
• CSRNet: Dilated Convolutional Neural Networks for Understanding the Highly Congested Scenes (CVPR2018) [paper] [code]
• Leveraging Unlabeled Data for Crowd Counting by Learning to Rank (CVPR2018) [paper] [code]
• Crowd Counting via Adversarial Cross-Scale Consistency Pursuit (CVPR2018) [paper]
• DecideNet: Counting Varying Density Crowds Through Attention Guided Detection and Density (CVPR2018) [paper]
• Crowd counting via scale-adaptive convolutional neural network (WACV2018) [paper] [code]

2017

• Generating High-Quality Crowd Density Maps using Contextual Pyramid CNNs (ICCV2017) [paper]
• Spatiotemporal Modeling for Crowd Counting in Videos (ICCV2017) [paper]
• CNN-based Cascaded Multi-task Learning of High-level Prior and Density Estimation for Crowd Counting (AVSS2017) [paper] [code]
• Switching Convolutional Neural Network for Crowd Counting (CVPR2017) [paper] [code]
• A Survey of Recent Advances in CNN-based Single Image Crowd Counting and Density Estimation (PR Letters) [paper]
• Image Crowd Counting Using Convolutional Neural Network and Markov Random Field (arXiv) [paper] [code]
• Multi-scale Convolution Neural Networks for Crowd Counting (arXiv) [paper] [code]

2016

• Towards perspective-free object counting with deep learning (ECCV2016) [paper] [code]
• Slicing Convolutional Neural Network for Crowd Video Understanding (CVPR2016) [paper] [code]
• CrowdNet: A Deep Convolutional Network for Dense Crowd Counting (CVPR2016) [paper] [code]
• Single-Image Crowd Counting via Multi-Column Convolutional Neural Network (CVPR2016) [paper] [code] [unofficial code]

2015

• COUNT Forest: CO-voting Uncertain Number of Targets using Random Forest for Crowd Density Estimation (ICCV2015) [paper]
• Cross-scene Crowd Counting via Deep Convolutional Neural Networks (CVPR2015) [paper] [code]

2013

• Multi-Source Multi-Scale Counting in Extremely Dense Crowd Images (CVPR2013) [paper]
• Crossing the Line: Crowd Counting by Integer Programming with Local Features (CVPR2013) [paper]

2012

• Feature mining for localised crowd counting (ECCV2012) [paper]

2008

• Privacy preserving crowd monitoring: Counting people without people models or tracking (CVPR 2008) [paper]

Datasets

• ShanghaiTech Dataset [Link: Dropbox / BaiduNetdisk]
• WorldExpo'10 Dataset [Link]
• UCF CC 50 Dataset [Link]
• Mall Dataset [Link]
• UCSD Dataset [Link]
• SmartCity Dataset [Link: GoogleDrive / BaiduNetdisk]
• AHU-Crowd Dataset [Link]

Performance

The section is being continually updated.

ShanghaiTech Part A

Method	MAE	MSE	PSNR	SSIM	Model Size	Params	Runtime (ms)	Pre-trained
DAN	81.8	134.7	-	-	-	-	-	-
CSR	68.2	115.0	23.79	0.76	-	-	-	-
L2R	73.6	112.0	-	-	-	-	-	-
ACSCP	75.7	102.7	-	-	5.1M	-	-	-
MCNN	110.2	173.2	21.4	0.52	0.12M	-	-	-

ShanghaiTech Part B

Method	MAE	MSE
DAN	13.2	20.1
BSAD	20.2	35.6
CSR	10.6	16.0
MCNN	26.4	41.3
L2R	13.7	21.4
DecideNet	21.53	31.98
ACSCP	17.2	27.4

UCF_CC_50

Method	MAE	MSE
DAN	309.6	402.64
BSAD	409.5	563.7
CSR	266.1	397.5
L2R	279.6	388.9
ACSCP	291.0	404.6

WorldExpo'10

Method	S1	S2	S3	S4	S5	Avg.
DAN	4.1	11.1	10.7	16.2	5.0	9.4
BSAD	4.1	21.7	11.9	11.0	3.5	10.5
CSR	2.9	11.5	8.6	16.6	3.4	8.6
DecideNet	2.0	13.14	8.90	17.40	4.75	9.23
ACSCP	2.8	14.05	9.6	8.1	2.9	7.5

UCSD

Method	MAE	MSE
BSAD	1.00	1.40
CSR	1.16	1.47
ACSCP	1.04	1.35

Tools

• Density Map Generation from Key Points [Code]