Automation Lab, Sungkyunkwan University

This is the official repository of

OpenCounting: An Open Source Implementation of Crowd Counting Methods.

1. Setup

1.1. Using environment.yml

conda env create -f environment.yml
conda activate anomaly

1.2. Using requirements.txt

conda create --name anomaly python=3.10.13
conda activate anomaly
pip install -r requirements.txt
pip install torch==1.13.0+cu117 torchvision==0.14.0+cu117 --extra-index-url https://download.pytorch.org/whl/cu117

2. Dataset Preparation

2.1. Image Crowd Counting Datasets

For the RGBT-CC dataset, please download it from this link.

For the ShanghaiTech RGB-D dataset, please download it from this repo.

For the UCF-QNRF dataset, please download it from this link

For the NWPU-Crowd dataset, please download it from this link

For the ShanghaiTech dataset, please download it from this link

For the STCrowd dataset, please download it from this link

For the NWPU-MOC dataset, please download it from this repository

For the Towards-vs-Away dataset, please download it from this repository

For the CARPK dataset, please download it from this link

For the FSC147 dataset, please download it from this link

For the FruitNeRF dataset, please download it from this link

For the StackCounting dataset, please download it from this link

For the TRANCOS dataset, please download it from this link

For the PUCPR dataset, please download it from this link

For the Crowd-SR dataset, please download it from this repository

For the Mall dataset, please download it from this repository

2.2. Video Crowd Counting Datasets

For the FDST dataset, please download it from this repository

For the VSCrowd dataset, please download it from this link

For the CroHD dataset, please download it from this link

For the CARLA dataset, please download it from this repository

For the MovingDroneCrowd dataset, please download it from this repository

For the DroneBird dataset, please download it from this repository

3. Usage

3.1 Supported Models for Bayesian Crowd Counting

Models	UCF-QNRF	ShanghaiTech
BayesianCrowd	✔️	✔️
NoisyCC	✔️	❌

3.1 Supported Models for Multimodal Crowd Counting

Models	RGBT-CC	ShanghaiTechRGBD
CSCA	✔️	✔️
IADM	✔️	✔️
EAEFNet	✔️	❌
MIANet	✔️	❌

3.2 Supported Models for VLM Crowd Counting

Models	ShanghaiTech	NWPU-Crowd	UCF-QNRF
CLIP-EBC	✔️	✔️	✔️

3.3 Supported Models for OT Crowd Counting

Models	ShanghaiTech	NWPU-Crowd	UCF-QNRF	Arbitrary Image
DM-Count	✔️	✔️	✔️	✔️
OT-M	❌	❌	❌	✔️
GeneralizedLoss	❌	❌	✔️	❌

3.4 Supported Models for INR Crowd Counting

Models	ShanghaiTech	NWPU-Crowd
APGCC	✔️	✔️
UNIC	✔️	❌
SI-INR	✔️	❌

3.5 Supported Models for Density Crowd Counting

Models	ShanghaiTech	FDST	UCF-QNRF	STCrowd	CARPK	Towards-vs-Away	Mall	JHU-Crowd++	NWPU-Crowd	TRANCOS	Crowd-SR
CSRNet	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
People-Flows	❌	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌
S-DCNet	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
SS-DCNet	✔️	❌	✔️	❌	❌	❌	❌	❌	❌	❌	❌
GCC-SFCN	❌	❌	✔️	❌	❌	❌	❌	❌	❌	❌	❌
CACC	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
SASNet	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
PAL	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
CUT	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
SGANet	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
RankBench	✔️	❌	✔️	❌	❌	❌	❌	❌	❌	❌	❌
STCrowd	❌	❌	❌	✔️	❌	❌	❌	❌	❌	❌	❌
FGCC	❌	❌	❌	❌	❌	✔️	❌	❌	❌	❌	❌
P2PNet	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
UEPNet	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
FIDTM	✔️	❌	✔️	❌	❌	❌	❌	✔️	✔️	✔️	❌
PML	✔️	❌	✔️	❌	❌	❌	❌	✔️	✔️	❌	❌
AutoScale	✔️	❌	✔️	❌	❌	❌	❌	✔️	✔️	❌	❌
IIM	✔️	✔️	✔️	❌	❌	❌	❌	✔️	✔️	❌	❌
DPD	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌
MSSRGN	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌	✔️
PMLoss	✔️	❌	❌	❌	❌	❌	❌	❌	❌	❌	❌

3.6 Supported Models for Domain Generalization Crowd Counting

Models	ShanghaiTech	UCF-QNRF
MPCount	✔️	✔️
DCCUS	✔️	❌
BLA	✔️	❌

3.7 Supported Models for Video Crowd Analysis

Models	SDD	IND-TIME	FDST	VSCROWD	JRDB	HT21	ETHUCY
CrowdMAC	✔️	✔️	✔️	✔️	✔️	✔️	✔️

3.8 Supported Models for Memory Bank Crowd Counting

Models	JHU-Crowd++
AWCC-Net	✔️

3.9 Supported Models for Transformer Crowd Counting

Models	JHU-Crowd++	NWPU	ShanghaiTech	UCF-QNRF	CARPK	NWPU-MOC	TRANCOS
CLTR	✔️	✔️	❌	❌	❌	❌	✔️
TransCrowd	❌	❌	✔️	✔️	❌	❌	❌
PET	❌	❌	✔️	❌	✔️	❌	❌
NWPU-MOC	❌	❌	❌	❌	❌	✔️	❌

3.9 Supported Models for MoE Crowd Counting

Models	ShanghaiTech
HMoDE	✔️

3.10 Supported Models for Knowledge Distillation Crowd Counting

Models	ShanghaiTech	UCF-QNRF	Mall
SKT	✔️	✔️	❌
LCSD	❌	❌	✔️
P2RLoss	✔️	❌	❌

3.11 Supported Models for Domain Adaptation Crowd Counting

Models	UCF-QNRF	Shanghaitech	CARPK	PUCPR
UGSDA	✔️	❌	❌	❌
CODA	❌	✔️	❌	❌
CBD	❌	❌	✔️	✔️

3.12 Supported Models for GCN Crowd Counting

Models	UCF-QNRF	JHU-Crowd++	ShanghaiTech
Gramformer	✔️	✔️	❌
GAAL	❌	❌	✔️
MDGCN	❌	❌	❌
DSGCNet	❌	❌	✔️

3.13 Supported Models for GCN Video Crowd Counting

Models	FDST
STGN	✔️

3.14 Supported Models for Open-world Object Counting

Models	ShanghaiTech	FSC147
OVID	✔️	✔️

3.15 Supported Models for Density Video Crowd Counting

Models	VSCrowd	CroHD	FDST	CARLA	MovingDroneCrowd	DroneBird
VSCrowd	✔️	❌	❌	❌	❌	❌
DAANet	✔️	✔️	❌	✔️	❌	❌
AVCC	❌	❌	✔️	❌	❌	❌
OMAN	✔️	❌	❌	❌	❌	❌
FMDC	❌	✔️	❌	✔️	❌	❌
CGNet	✔️	❌	❌	❌	❌	❌
DRNet	✔️	✔️	❌	❌	❌	❌
MovingDroneCrowd	❌	❌	❌	❌	✔️	❌
DroneBird	❌	❌	❌	❌	❌	✔️

3.16 Supported Models for Low-shot Crowd Counting

Models	FSC147
FamNet	✔️

3.17 Supported Models for 3D Crowd Counting

Models	FruitNeRF	StackCounting
FruitNeRF	✔️	❌
FruitNeRF++	✔️	❌
3DC	❌	✔️

4. Citation

If you find our work useful, please cite the following:

@misc{Chi2023,
  author       = {Chi Tran},
  title        = {OpenCrowd: An Open Source Implementation of Crowd Counting Methods},
  publisher    = {GitHub},
  booktitle    = {GitHub repository},
  howpublished = {https://github.com/SKKU-AutoLab-VSW/ETSS-05-CongestionDetection},
  year         = {2025}
}

5. Contact

If you have any questions, feel free to contact Chi Tran (ctran743@gmail.com).

6. Acknowledgement

Our framework is built using multiple open source, thanks for their great contributions.