question about Gaussion kernel size and CAR useage during training ROMP

[ChangjianLi]

作者大大你好 I have a question about the training codes. I have read through the codes and I couldn't find the methods of Gaussion kernel size and CAR calculation used for the training. Did I missed some part lol

[Arthur151]

Hi, The implementation is quite simple CAR: https://github.com/Arthur151/ROMP/blob/91dac0172c4dc0685b97f96eda9a3a53c626da47/romp/lib/maps_utils/centermap.py#L99 Gaussion kernel size: https://github.com/Arthur151/ROMP/blob/91dac0172c4dc0685b97f96eda9a3a53c626da47/romp/lib/maps_utils/centermap.py#L362

[ChangjianLi]

Hi, The implementation is quite simple CAR:

https://github.com/Arthur151/ROMP/blob/91dac0172c4dc0685b97f96eda9a3a53c626da47/romp/lib/maps_utils/centermap.py#L99

Gaussion kernel size: https://github.com/Arthur151/ROMP/blob/91dac0172c4dc0685b97f96eda9a3a53c626da47/romp/lib/maps_utils/centermap.py#L362

Thx for the reply : ). But are these methods used during training and testing? I looked through the result_parser.py and the function used to calculate centermap is self.centermap_parser.parse_centermap(outputs['center_map']) instead of CAR method

[Arthur151]

They are just used to generate the ground truth body center heatmap to guide more discriminative representation learning. In other word, they are only used for training.