Project "Mechanische Integration WS 2021/22". Counting Pedestrians using computer vision on a Nvidia Jetson Nano 4GB.
For the course "Mechanische Integration WS 21/22" held by Prof. Dr. Leibl and Prof. Dr. Steinkogler in cooperation with the "Stadt München Referat Mobilität" our group developed a pedestrian counting system using computer vision.
Features include:
- Webserver for setup and control
- Support for different Tracking Algorithms (cv2.CSRT, cv2.KCF, dlib correlation Tracker)
- MobileNetSSD using Caffe Framework
- Centroid Tracking ID algorithm
- Ouput of CSV files with number and direction of detected Pedestrians
- GPU acceleration on supported Hardware
- Optimized performance by threading of workload
output_gud.video-converter.com.mp4
Code by: Luis Rebhan and Jakub Zelonka