Nicolaj Schmid,Danial Zendehdel, Ekrem Yüksel, Thomas Jaouën
Except that Loomos can provide mobility service, and among all of its features, it allows the user to deploy its model and test it on a real-life gadget.
This project aims to detect a person as the person of interest, which has been done here by the Openpifpaf algorithm and then Track the recognized person by the yolov4-deepsort algorithm.
Openpifpaf is used In order to recognise a person as a target for the tracking part. This is done by introducing a pose rule in which the left wrist is higher than left shoulder and the right wrist is lower than right shoulder.
As the loomos work with low resolution the images are downscaled to 160*120.
The Tracking is based on the Yolovv4-Deepsort, robust in low-resolution images and Tested on V100.
zzz.mov
To get started, install the proper dependencies via Pip.
$ virtualenv <env_name>
$ source <env_name>/bin/activate
(<env_name>)$ pip install -r path/to/requirements.txtInstall openpifpaf
pip3 install openpifpafFor images
python openpifpaf_api.py --datapath data/test_img2.png python client.py --ip-address 128.179.159.152 --d 4