Three demo codes.

demo.py: uses only one instance of PygoTurn

demo_fb.py: uses two instances, that move in forward and backward directions, respectively.

demo_delta.py: uses two instances both moving in the forward direction, separated delta frames.

Navigate to pygoturn/src and do:

python3 demo.py -w /path/to/pretrained/model

Arguments:

-w / --model-weights: Path to a PyTorch pretrained model checkpoint.

-d / --data-directory: Path to a tracking sequence which follows OTB format.

-s / --save-directory: Directory to save sequence images with predicted bounding boxes.

-method : method used for combining the bounding boxes of the two instances. iou: chose the BB with max IoU (uses the ground truth). mean: computes the mean between the two BBs (no grund truth). By default we just use one of the instances.

-plots : directory where to save the numpy arrays of IoU for plotting.

Example:

python3 PygoTurn/PygoTurn\ Implementation/src/demo_delta.py -w pytorch_goturn.pth.tar -d PygoTurn/PygoTurn\ Implementation/sequences-train/book/ -s PygoTurn/PygoTurn\ Implementation/our_results/book/ -plots PygoTurn/PygoTurn\ Implementation/our_results/plots/book -method mean