The question about filter_db function.

[wanghangege]

why use the function (filter_db) to filter the difficulty and min_points ? Does it make bad influence to the performance?

def filter_db(self, db_infos):
    # 1. filter_by_difficulty
    for k, v in db_infos.items():
        db_infos[k] = [item for item in v if item['difficulty'] != -1]

    # 2. filter_by_min_points, dict(Car=5, Pedestrian=10, Cyclist=10)
    filter_thrs = dict(Car=5, Pedestrian=10, Cyclist=10)
    for cat in self.CLASSES:
        filter_thr = filter_thrs[cat]
        db_infos[cat] = [item for item in db_infos[cat] if item['num_points_in_gt'] >= filter_thr]

    return db_infos

[zhulf0804]

1. Difficulty value is calculate here https://github.com/zhulf0804/PointPillars/blob/b9948e73505c8d6bfa631ffdf76c7148e82c5942/pre_process_kitti.py#L16-L32 And objects with difficulty = -1 are not cared as described in https://www.cvlibs.net/datasets/kitti/eval_object.php?obj_benchmark=3d
2. Filtering by min_points may influence the performance. We keep consistent with filter_by_min_points in mmdet3d. Also the parameters are updated for a better performance in pr 1166

[jonasdieker]

Hey @zhulf0804, do I understand this correctly:

The filtering here is only for the annotations database, which is used for the data augmentation when sampling extra ground truth bboxes and points. You do not filter out difficult bboxes from the actual ground truths of a particular frame when loading the data, right?