ML expert needed for Multi- person-tracking and re-identification algorithm to identify whether it is the same person in the two videos in the different angle -- 2
Budget: $30 – $200 USD
Note that a person detected in one camera frame will be labelled the same in another camera frame too. It's not to simply detect all people in the group. To read more about the requirement read this paper:
http://www.diva-portal.org/smash/get/diva2:1395007/FULLTEXT01.pdf
Also, I have attached one pdf
The end result will look like this: it will give you the correct direction
https://youtu.be/sroLfpX4F0w
another example of the same result :
https://youtu.be/YhW7YRJvQpg
The example code is here:
https://github.com/Mhttx2016/Multi-Camera-Object-Tracking-via-Transferring-Representation-to-Top-View
The above code is just an example, they calibrate the 4 videos. they use some pickle files for each video.
But I want you to use use yolo model for detection instead of the pickle which they are using and furthermore, We will be having only two videos as a input. so the logic changes completely.
Note, above github link is only for reference. We have tried using that approach but couldn't get the result.
So you have find your own way of doing this.
Furthermore , two videos will be in overlapping view. so the same person appears at the same time in different angle
you can use the following dataset to generate output: https://drive.google.com/drive/folders/1eFy5D9kUSDoarnjyvZUFghdhhBu14Kor
New freelancers are also welcomed, only grit & perseverance is required.
http://www.diva-portal.org/smash/get/diva2:1395007/FULLTEXT01.pdf
Also, I have attached one pdf
The end result will look like this: it will give you the correct direction
https://youtu.be/sroLfpX4F0w
another example of the same result :
https://youtu.be/YhW7YRJvQpg
The example code is here:
https://github.com/Mhttx2016/Multi-Camera-Object-Tracking-via-Transferring-Representation-to-Top-View
The above code is just an example, they calibrate the 4 videos. they use some pickle files for each video.
But I want you to use use yolo model for detection instead of the pickle which they are using and furthermore, We will be having only two videos as a input. so the logic changes completely.
Note, above github link is only for reference. We have tried using that approach but couldn't get the result.
So you have find your own way of doing this.
Furthermore , two videos will be in overlapping view. so the same person appears at the same time in different angle
you can use the following dataset to generate output: https://drive.google.com/drive/folders/1eFy5D9kUSDoarnjyvZUFghdhhBu14Kor
New freelancers are also welcomed, only grit & perseverance is required.