This is a packaging implementation of the paper CNN-based Cascaded Multi-task Learning of High-level Prior and Density Estimation for Crowd Counting for single image crowd counting which is accepted at AVSS 2017
The package is compatible with all operating systems, provides a staggering fast and accurate prediction. It achieves a min of 20 fps on a 6 core intel cpu.
pip install ezcrowdcount
To run inference on your favorite image/video simply run the following on your terminal/console:
crowdcount --mode video --path /path/to/video
"""
mode (str): Whether to run prediction on video or image
path (str | int): Path to video or image. It can be an index to a camera feed, or a URL also. (Default = 0).
"""
The inference will run on your GPU (if available), and will be viewed right in front of you 👀 Also, the number of people during each frame will be printed on your console/terminal.
Input Image:
Result Image: