This is the implementation of the object tracker, GOTURN: Generic Object Tracking Using Regression Networks, written in Julia language using Knet: Koc University Deep Learning Framework.
Paper: https://arxiv.org/pdf/1604.01802.pdf
Needed datasets: For training, VOT2014. For testing ALOV300+.
Arrangement of the dataset folders should be as the following:
|-goturn.jl
|-datasets
|---trn
|-----annotations
|-----images
|---tst
|-------(video folders)
|---val
|-----annotations
|-----images
Check inside main function for arguments. Default arguments provide best tested results.