Hao Shao

The more the better, `routes_training.xml` has the most proportion of the data.

No, we collect the drivinf data based on other xmls (please refer to repo#data collection)

Hi! We just submit our model to the leaderboard and get the results reported in our paper. We also don't have more details, and maybe the official repo `leaderboard` can...

Hi! > When we collect data, we run the leaderboard_ evaluator.py, this process will also result in the evaluation of the data, right? Yes > If I only performed data...

Hi! > In the Waypoint loss function, the waypoints generated by the expert agent are the waypoints used during data collection. Is that right？ No, the ground truth waypoints when...

Hi! We also found this problem in our project. You can try to lower the predicted prob in the corrsponding area of the object density map.

Hi! Most of them have reported the corrsponding results in these benchmarks, we just copy them to here. We retrain the others models with the data we collected. Because different...

典型错误，确保kill掉所有carla的进程，再重启试试。不行再试试别的端口

Hi! Our repo provides this function: https://github.com/opendilab/InterFuser/blob/e4f0314482124bb06a475c3f6fb4bfe3a2701c4d/scenario_runner/srunner/scenariomanager/carla_data_provider.py#L759 Please export out `scenario_runner` as `SCENARIO_RUNNER` path

Hi! About these scenarios, we refer to the Transfuser repo. You may find more details in their repo and paper.