The block is used to benchmark any tracking algorithm against the nuScenes dataset.

The block is used to append all the algorithm's predictions before/after the preprocessing process to be evaluated or submitted later. It saves the results as a json file with the same format of the nuScenes Challenges. It works for the detection/tracking evaluation. It will discard the boxes with no lidar and radar points.

The block is used to load the results of the predictions from a json file. It can be a detection or tracking results.

The block is used to perform preprocessing on the eval boxes either for the detection or tracking evaluation. First, add the distance between the boxes to the ego vehicle. Second, filter the boxes according to the calculated distance, number of lidar/radar points in the boxes, and either the object is a bicycle in the racks or not.

The block is used to convert the nuScenes 3D bounding boxes in the sample annotation format to evaluation format.