DeepSort NN - feature request

[trackme518]

Hi,
first of all thanks for the awesome library! I am testing it right now and it is very nice to see YOLO4tiny implemented. However, I am into tracking stuff and while the YOLO could be implemented into a more traditional blob tracking pipeline I would like to see a proper DeepSORT implementation.

Please have a look at this: https://github.com/theAIGuysCode/yolov4-deepsort

I have already run the DeepSort from that repository on separate computers and it works very well for tracking purposes. As the DeepSort is just YOLO+SORT it might not be that hard to implement since you have YOLO already and the linked GitHub repo is using OpenCV DNN as well.

Let me know if you feel like doing it and if there is something I can assist you with.

[cansik]

DeepSORT looks interesting, I am mainly using motpy for multi-object tracking using a more traditional approach. I will have a look if and how it could be integrated. If you want to contribute, your are of course more than welcome to create a PR.

But to be honest, I started writing a new computer vision framework (similar to deep-vision) for python, because it is so much easier to integrate recent development in that field into an existing python pipeline, instead of rewriting everything in java. Also some frameworks are way better integrated, for example MediaPipe or OpenVINO. The framework is not public yet, but I am using it already in production for exhibitions and performances.

[trackme518]

Hi, thanks for fast reply. Yeah java is a bit cumbersome in that regard i know, i just happened to use processing a lot for past projects so it make sense to extend it in this area. Any chance you would be releasing the python framework you mentioned? Also I am working on spatial audio right now in dresden htw and i also do PhD at cvut in Prague - if you are into publishing paper i think we can collaborate on some projects. I also have a job as researcher at lasvit.com - large scale immersive installations using light and hand blown glass. Mar 17, 2022 14:51:12 Florian Bruggisser ***@***.***>:

…

DeepSORT looks interesting, I am mainly using motpy[https://github.com/wmuron/motpy] for multi-object tracking using a more traditional approach. I will have a look if and how it could be integrated. If you want to contribute, your are of course more than welcome to create a PR. But to be honest, I started writing a new computer vision framework (similar to deep-vision) for python, because it is so much easier to integrate recent development in that field into an existing python pipeline, instead of rewriting everything in java. Also some frameworks are way better integrated, for example MediaPipe or OpenVINO. The framework is not public yet, but I am using it already in production for exhibitions and performances. — Reply to this email directly, view it on GitHub[#13 (comment)], or unsubscribe[https://github.com/notifications/unsubscribe-auth/AL7PGZJH6C3IN5HQLVVBVKDVAM2E3ANCNFSM5Q67DUOQ]. Triage notifications on the go with GitHub Mobile for iOS[https://apps.apple.com/app/apple-store/id1477376905?ct=notification-email&mt=8&pt=524675] or Android[https://play.google.com/store/apps/details?id=com.github.android&referrer=utm_campaign%3Dnotification-email%26utm_medium%3Demail%26utm_source%3Dgithub]. You are receiving this because you authored the thread.[Tracking image][https://github.com/notifications/beacon/AL7PGZNOGTNSFNZDA5FH2B3VAM2E3A5CNFSM5Q67DUO2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOH7KWGLQ.gif]