PY-GOTURN
Update
There wont be any new updates to this repository as official Caffe is no longer maintained. But as long as you compile Caffe successfully, this repository should work as is.
Instead, try PyTorch version here.
Some of the others' work using this repository:
NOTE: Please switch to goturn-0.1 branch to try out stable version.
This is the python + Caffe implementation of GOTURN: Generic Object Tracking Using Regression Networks.
Learning to Track at 100 FPS with Deep Regression Networks,
David Held,
Sebastian Thrun,
Silvio Savarese,
Outputs
| Car | Sunshade |
|---|---|
 |
 |
Why implementation in python, when C++ code is already available?
- Easy to understand the overall pipeline of the algorithm in detail
- Easy to experiment new ideas
- Easy to debug and visualize the network with tools like visdom
- Little effort in portability to other OS
Functionalites added so far
-
Training the deep network on Imagenet and ALOV dataset
-
Test code for VOT
How to train your own tracker
To train your own tracker, please follow the guide
How to test on VOT dataset
To test your own tracker on VOT2014, please follow the guide
Contact
Please write to me at [email protected], if you have any suggestions or any new functionality you like to see.