Viktor Kocur

Version of the file from an [older commit](https://github.com/kocurvik/retinanet_traffic_3D/blob/492533a1fb3c97d70efbfd23f336ced836d18436/dataset_utils/fromBrnoCompSpeed.py) should work.

It should work with OpenCV 4.2.0, but there might be some other issue.

Can you check the variables that are input into the function? Maybe there was an error and one of them is a `None` or something.

Hi, I do not own the dataset. Please contact Vojtěch Bartl from Graph@FIT - [email protected] or Jakub Špaňhel - [email protected]

The results from the paper are provided in another [repository](https://github.com/kocurvik/BCS_results). You can also obtain them by running dataset_utils/test.py. To evaluate the results please refer to the official BrnoCompSpeed evaluation code...

Hi, the tracking data gets saved to a file defined here: https://github.com/kocurvik/retinanet_traffic_3D/blob/592ceac767750c65bb3d6678b36e6880a7bb0403/dataset_utils/tracker.py#L61 That is the results folder of the BrnoCompSpeed dataset. The name is usually defined by the model name...

> HI, I have a question. If I want to do speed test on BrnoCompSpeed dataset, must I do track_detections function in dataset_utils/test.py to get 'system_{write_name}.json'? @jinqi2376 Have you successfully...

You are not supposed to use test_luvizon at all if you are working with BrnoCompSpeed. Once you have the file `system_name_of_the_system.json` you have to use the [official BrnoCompSpeed evaluation code](https://github.com/JakubSochor/BrnoCompSpeed)...

> Could you try running `scripts/build_python_package.sh`? I have seen the pull request even before getting merged and it did not work. I have tried it now again. There are still...

I created PR #98 with the fix, but you are right that it does not actually affect performance.