search

McMasterAI / TrafficTracker

A street traffic counter that provides efficient and ethical modelling of how people engage and interact within a given public space.
MIT License
7 stars 2 forks source link

Investigate Implementation of YOLOv5 #26

Closed cczarnuch closed 4 years ago

cczarnuch commented 4 years ago

Investigate the implementation of YOLOv4 in the pythonlessons repo and how we can replicate that with YOLOv5.

cczarnuch commented 4 years ago

YOLO/R-CNN comparison: https://towardsdatascience.com/yolov5-compared-to-faster-rcnn-who-wins-a771cd6c9fb4

YOLOv5 repo: https://github.com/ultralytics/yolov5

cczarnuch commented 4 years ago

If I'm not mistaken, it looks like we can pull in the 'utils' and 'models' folders from the yolov5 repo and that will be enough to run the detections. We should use the release version and not the repo itself to do this. we can put it all in a yolov5 package like in the python lessons repo.

Once we have that, we can re-create our own 'detect.py' script to run the detections which will be used in the deep_sort.py app which will combine everything.

A couple things that came to mind, we may need to create our own model (labelling images) etc.

It would be very helpful if we could get some documentation once this story is complete outlining your findings.

@djurdjm @icecap360