Closed WillJStone closed 5 years ago
.pyc
files should probably not be commited -- maybe add them to a .gitignore
file?
.pyc
files should probably not be commited -- maybe add them to a.gitignore
file?
I will do this, thanks!
Ideally, we should "link" to the other repo using git submodules.
What were the files that you coded by yourself?
Ideally, we should "link" to the other repo using git submodules.
What were the files that you coded by yourself?
The only new file that I coded myself is dasiamrpn.py. I figured there was a better way to do this than just include the other repo but was unaware of how to do it. I will look into the submodule functionality.
I've made the requested changes. I have also added a file containing the pre-trained weights for the DaSiamRPN network. It's 86 MB so I'm not sure if it should be part of the repo or not.
Hello, great! Did you have a look at using git submodules?
I think 90 MB it is ok.
Great code!
I got the following error:
I can fix it by adding a .
before utils
(line 11, DaSiamRPN/code/run_SiamRPN.py), however, this will not come like this when we clone the submodel.
Do you have any other idea of how we could fix this?
Hello @WillJStone, did you manage to do the fork of the DaSiamRPN repo and make the previously mentioned changes?
It's on my (very long) list of things to do. I hope to get around to it this weekend. I have figured out exactly what needs doing, just a matter of doing it. Is there time pressure involving another pull request you want to merge that may conflict?
I fixed the previously discussed issues (:
Thank you for this PR!
I have implemented a wrapper class for the DaSiamRPN tracking network (https://github.com/foolwood/DaSiamRPN) such that it can be used instead of the OpenCV tracking algorithms that can currently be used in the OpenLabeling software. This method of tracking significantly out-performs those currently available via OpenCV.
I have also added an --n_frames flag to the argparser. This represents the number of frames that the tracking method will operate for before terminating. I have noticed that the OpenCV trackers are faulty on the best of days with regards to realizing when they have failed, and I couldn't think of an effective method of deciding when the DaSiamRPN method has failed. To solve this, I figured it was just better for the user to determine how many frames they think the algorithm can go for given their specific data.
PS: I have never submitted a PR to anything before so please let me know if there is anything I've set up here that isn't optimal, looking to learn!