the confidence threshold you set for the pretrained detector

[linhaojia13]

As you mentioned in the papers:

... Group-Free detector [33] for 3D point clouds pre-trained on a vocabulary of 485 object categories on ScanNet [8]. The detected box proposals that surpass a confidence threshold are encoded using a box proposal encoder ...

Could you please tell me the confidence threshold you set for the pretrained Group-Free detector?

[ayushjain1144]

This is what we used:

eval_config_dict = {'remove_empty_box': True, 'use_3d_nms': True, 'nms_iou': 0.25,
'use_old_type_nms': False, 'cls_nms': False, 'per_class_proposal': False,
'conf_thresh': 0.5, 'dataset_config': DC}

Note that we didn't really play around with these, so they may not be optimal.

[linhaojia13]

This is what we used:

eval_config_dict = {'remove_empty_box': True, 'use_3d_nms': True, 'nms_iou': 0.25,
'use_old_type_nms': False, 'cls_nms': False, 'per_class_proposal': False,
'conf_thresh': 0.5, 'dataset_config': DC}

Note that we didn't really play around with these, so they may not be optimal.

Thank you very much！