How to evaluate the performance of the proposed method on the KITTI dataset？ about grnet HOT 6 CLOSED

The above error occurred due to the lack of ground truth of 3D objects.

The code for Uniformity can be found in https://github.com/liruihui/PU-GAN.
The code for Consistency can be implemented by yourself. It is easy, isn't it?

You can generate the prediction for KITTI with the following commands:

python runner.py --inference --weights /path/to/pretrained/model

from grnet.

The above error occurred due to the lack of ground truth of 3D objects.

The code for Uniformity can be found in https://github.com/liruihui/PU-GAN.
The code for Consistency can be implemented by yourself. It is easy, isn't it?

You can generate the prediction for KITTI with the following commands:

python runner.py --inference --weights /path/to/pretrained/model

OK, thank you very much！

from grnet.

Hi @zhangyanan0306 Did you successfully run the uniformity evaluation? The code (https://github.com/liruihui/PU-GAN) requires a mesh input, but now we only have point cloud results as well as point gt.

@hzxie Would you please share the uniformity code for testing on KITTI? Thanks!

from grnet.

@zhangmozhe
Please refer to liruihui/PU-GAN#18

from grnet.

@zhangmozhe
Please refer to liruihui/PU-GAN#18

Their code requires mesh as input. Then how to compute the metric if we only have the point cloud ground truth? Thanks.

from grnet.

The code for uniformity is available in #20.

from grnet.