Semi-Supervised Crowd Counting with Contextual Modeling: Facilitating Holistic Understanding of Crowd Scenes
The list of images in the validation set for UCF-QNRF is the same as that in BL. Please set the learning rate to 2e-5, weight decay to 4e-5 when training on QNRF.
Some more details will be provided later
The pretrained models could be downloaded from Google Drive or OneDrive. Please test the pre-trained model on the images pre-processed with our code. We just realize the results would siginificantly change if the quality of the images changes for QNRF.
timm==0.5.4
python < 3.10
pytorch >=1.4
opencv-python
scipy==1.6.2
h5py
pillow
tqdm
If you find our work useful, please cite:
@ARTICLE{Semi2024Qian,
author={Qian, Yifei and Hong, Xiaopeng and Guo, Zhongliang and Arandjelović, Ognjen and Donovan, Carl R.},
journal={IEEE Transactions on Circuits and Systems for Video Technology},
title={Semi-Supervised Crowd Counting With Contextual Modeling: Facilitating Holistic Understanding of Crowd Scenes},
year={2024},
volume={34},
number={9},
pages={8230-8241},
doi={10.1109/TCSVT.2024.3392500}}