validation set gt bboxes

[vutrungtam619]

Hi, thanks for this project. I want to ask about kitti.py, as I see, in train set you using object_range_filter, but why in val set you dont use it? It's mean you still calculate loss on gt box out of point cloud range?

        if self.split in ['train', 'trainval']:
            data_dict = data_augment(self.CLASSES, self.data_root, data_dict, self.data_aug_config)
        else:
            data_dict = point_range_filter(data_dict, point_range=self.data_aug_config['point_range_filter'])