Crowd counting through drones using VisDrone 2020/2021 Dataset & Neural Networks via TransCrowd transformer
The VisDrone Dataset could be found here.
- Task 5: Crowd Counting
The TransCrowd implementation can be found under the directory "/TransCrowd".
For official implementation, have a look here.
You can run the whole process via the Google Colab Notebook that can be found here.
It is highly suggested to run the notebook on a Pro account, since loading VisDrone image list requires at least 16GB of RAM (Colab Free gives 12GB).
- Gianmarco Scarano (Owner)
- Gennaro Vessio (Supervisor)
- Pasquale De Marinis (Test annotations converter)
- Credits for TransCrowd implementation:
@article{liang2021transcrowd,
title={TransCrowd: Weakly-Supervised Crowd Counting with Transformer},
author={Liang, Dingkang and Chen, Xiwu and Xu, Wei and Zhou, Yu and Bai, Xiang},
journal={arXiv preprint arXiv:2104.09116},
year={2021}
}
