You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
The only problem here is that we need either to support MinkowskiEngine in mmdetection3d or rewrite the model to spconv backend (better after updating it to v2.* as v1.* is not supported now). Now v1.* is migrating to mmcv in this PR. This rewriting is relatively simple but then it will need some time to achieve same metrics as with MinkowskiEngine backend.
Hi @fcakyon Thanks for your suggestion. As @filaPro pointed out, the requirement of Mink Engine needs some work for us to support this awesome method. We will consider it carefully after finishing migrating spconv to mmcv, although there is still a lot of work about it (such as refactoring it and upgrading it to fit our framework) that need to be done.
Do yo have any plans to add FCAF3D support which is already implemented with mmdet3d source? https://github.com/samsunglabs/fcaf3d
The text was updated successfully, but these errors were encountered: