This code is a fork of https://github.com/eriklindernoren/PyTorch-YOLOv3. Here, this repo is modified to incorporate tracking of detected elements and output an OD-data formatted CSV.
Installation
Clone this repo, and install the requierements (you may want to use a virtual pip env for that). For faster computation of model's inferences, you can install CUDA.
$ sudo pip install -r requirements.txt
Run
For the tracking, need to use the script dectect.py as it follows:
$ python3 detect.py --image_folder data/samples/
Any folder can be used, data/samples/ is simply the default.
The tracking rely on the hypothesis that images in your folder names are incremental (or at least sorted).