- Install dependencies using conda from
environment.yml
- Create
autonomous-traffic-monitoring/external/
folder - Install Detectron2
in
autonomous-traffic-monitoring/external/
- Download and add Mask R-CNN model weights (R101-FPN)
from Detectron2 Model Zoo
in
autonomous-traffic-monitoring/detection/detetctron2-model-weights
- Clone DeepSORT repository and put it in
autonomous-traffic-monitoring/external/
- Clone IOU/VIOU Tracker repository and put it in
autonomous-traffic-monitoring/external/
- Download and add DeepSORT model weights
in
autonomous-traffic-monitoring/tracking/deep-sort-model-weights/
- [OPTIONAL (alternative camera calibration)] Install monodepth2 in and put
model
weights (mono+stereo_640x192)
in
autonomous-traffic-monitoring/camera/models/
auto-traffic-monitoring/input/scripts/get_tfl_videos.py
auto-traffic-monitoring/input/scripts/get_frames_from_video.py video.mp4
Get landmarks from location of camera in Google Maps and match points with frame.
Format .csv
file as u, v, X, Y, Y
- Main method (recommended):
auto-traffic-monitoring/camera/run_camera.py --video_path input/videos/video.mp4 --landmarks_path input/landmarks/video_landmarks.csv --mode manual
- Alternative method:
auto-traffic-monitoring/camera/run_camera.py --video_path input/videos/video.mp4 --landmarks_path input/landmarks/video_landmarks.csv --mode manual
auto-traffic-monitoring/detection/run_detection.py --video_path input/videos/video.mp4 --camera_params results/video.mp4_camera_parameters.pickle
auto-traffic-monitoring/tracking/run_tracking.py --video_path input/videos/video.mp4 --detection results/video.mp4_detection.csv --camera_params results/video.mp4_camera_parameters.pickle
auto-traffic-monitoring/tracking/run_speed_estimation.py --video_path input/videos/video.mp4 --tracking results/video.mp4_tracking_iou.csv --camera_params results/video.mp4_camera_parameters.pickle