Yukai Yang (Alexis)

I suggest you try using the homography matrix below directly and `utils.numba.perspective_transform()` to warp the end points of the line. https://github.com/GeekAlexis/FastMOT/blob/8f59591aa10adaab84af7c083bc98d1071a56bd8/fastmot/flow.py#L223

I don't have plans to port the repo to C++. There is more room for multithreading with C++ for sure but I doubt C++ can provide a huge FPS boost...

@pductruong Hi, the repo only supports CUDA at the moment. I will consider adding a backend with onnxruntime to support both CPU and GPU.

Any update on this? I got off bounding boxes using the c++ infer api in the master branch as well with 960x540 resolution.

Hi have you tried to install on Jetpack 5? You can comment out the TensorFlow installation for now: https://github.com/GeekAlexis/FastMOT/blob/94c10a669485d138267b633e30ca1e577f2b1a21/scripts/install_jetson.sh#L37 Since different Jetpack versions require specific TensorFlow versions. Also pull requests...

> I managed to get it to work way back but forgot to reply here, however I'm fairly new to github so what are pull requests? Thanks for your update....

Make sure you set the network parameters in `fastmot.models` correctly for your model.

Hi guys, if you are looking for a real-time tracker with TensorRT acceleration, here is my implementation: https://github.com/GeekAlexis/FastMOT, which is more practical than FairMOT at the moment. Please star the...

Hi guys, if you are looking for a highly optimized Deep SORT and YOLOv4 with TensorRT acceleration, here is my implementation: https://github.com/GeekAlexis/FastMOT. Let me know if you can get it...

I have the same issue. It looks like `train_loop` doesn't return the loss at all.