Tai Wang comments

Results 118 comments of


                                            Tai Wang

python setup.py develop failed

It's hard to say what is the exact reason triggering your compilation error, but I notice your Cuda version is inconsistent among nvcc, pytorch and mmcv. Maybe you can double-check...

Sparse shape causing problems

This bug seems not directly related to the setting of sparse_shape. Could you please show your entire config?

These codes maybe meaningless and can cause an error in certain circumstances

Thanks for your kind reminder. Could you please create a PR to make it more generalizable?

How to configure the environment and start training

Please follow the issue template for "reimplementation" to provide more information about your case.

There is no implementation of 'sample_2d' in ObjectSample

Please refer to the DBSampler in this [PR](https://github.com/open-mmlab/mmdetection3d/pull/327).

Pretrained model could not detect 3d objects in my own point cloud data

Maybe you can compare your converted data with our provided demo data? It might be related to the different coordinate systems or points' coordinate scales?

关于Reszie3D

请问您用的哪个config？看起来不像是repo里给定的。FCOS3D或者PGD在KITTI上的Resize3D如何使用还有待测试，如有需要可以先用DfM和Waymo上PGD的config尝试。

Indoor object detection on the way?

Please refer to [ImVoxelNet](https://github.com/saic-vul/imvoxelnet) and [mmdet3d](https://github.com/open-mmlab/mmdetection3d) for some basic indoor methods. Further exploration in this direction is ongoing and please stay tuned.

[Enhance] Remove CUDA part from diff_iou_rotated

LGTM. Just need to resolve conflicts