In this repository, a deep learning based approach for surveillance camera is presented. Our system composes of 3 tasks:
- Object Detection: Automatically localizing and classifying different types of vehicles on the street. This is done using YOLOv4-tiny [1].
- Object Tracking: Assign ID to each detected object. The object tracker is applied using Deep Sort [2].
- Lane detection: Classifying each pixel if it belongs to a lane line or not. This is done using a U-net based segmentation neural network. [3].
- Train YOLOv4-tiny using Darknet repository
- Run save_model.py
- Train U-net using \ref{} this repo
- Run DBSSIT.py with saved YOLO and U-net models.
[1] Wang, C.-Y., Bochkovskiy, A., and Liao, H.-Y. M., “Scaled-YOLOv4: Scaling Cross Stage Partial Network”, arXiv e-prints, 2020.
[2] N. Wojke, A. Bewley and D. Paulus, "Simple online and realtime tracking with a deep association metric," 2017 IEEE International Conference on Image Processing (ICIP), 2017, pp. 3645-3649, doi: 10.1109/ICIP.2017.8296962.
[3] Ronneberger, O., Fischer, P., and Brox, T., “U-Net: Convolutional Networks for Biomedical Image Segmentation”, arXiv e-prints, 2015.