auron-x / goturn_training_toolkit Goto Github PK

when I g++ buildH5Datasets.cpp:
undefined reference to 'cv::gtr::anno2rect(std::vector<cv::Point_, std::allocator<cv::Point_ > >)'

undefined reference to 'cv::gtr::gatherFrameSamples(cv::Mat, cv::Mat, cv::Rect_,cv::Rect_)'

How can I solve this problem?

Hello,

I am confused about the role the searching image plays in the training of GOTURN.

The bounding box points that are provided are for the training image correct?

I would like to train on my own videos. Say i have a video of a car driving by, then say i annotate each frame of the video with a bounding box. How do i then go about organizing the frames into the searching image directory, and the training image directory? Do i even have to annotate every frame in the video?

Thanks

I have read the code ，but I still no clear about the process of generate goturn.prototxt and goturn.caffemodel

Hi, I have trained this model with GPU 1080TI for 2days, and the loss is about 400. I want to know is it normal?