:integrate to VOT benchmark:

[sydney0zq]

No description provided.

[nrupatunga]

Hi @sydney0zq thank you for you interest in contributing to the project,
Can you please briefly describe your changes and also provide the instructions to test it
Step by step instruction would be much appreciated .
Thank you

[sydney0zq]

How to test our model on VOT benchmark?

1. Firstly, we need modify this script properly to refer several necessary resource files, such as caffemodel and prototxt file etc, as well as some environment variables
2. Secondly, clone vot-toolkit repo and set tracker_GOTURN.m associated to our script
3. Thirdly, follow VOT official documentation to test our tracker on VOT dataset

As a reference, my EAO on VOT2015 is about 20.4

[sydney0zq]

By the way, we need change tracker object. I think our regressor should be consistent over one single video. I will pull another request soon.

[sydney0zq]

Here notice I change the class from goturn.tracker.vot_tracker import tracker.

[devyhia]

@sydney0zq Did you test a trained model against the VOT challenge? If so, could you please share the result as well?

[sydney0zq]

@devyhia No, I just use origin repo’s pretrained model on VOT2015.

[devyhia]

@sydney0zq Okay, I will use this pull request to test the trained model against the VOT challenge then. I hope it yields close results to those Held achieved.