Renrui Zhang
Renrui Zhang
Sorry for confusion. Our paper is still under reviewed. :)
@si22bao Thanks for your waiting. A more stable code of MonoDETR has been released.
Thanks for your interest. The nuscenes results are under tuned now and we'll share them soon!
Thanks for your interest! We directly transfer the code of PointCLIP into the code base of different methods, e.g. SimpleView's and CurveNet's. More simply, you can cache the predicted logits...
Sorry for the delay! We have released the pre-training code. The fine-tuning code will be available in a few days.
Sorry for making you wait. The fine-tuning code of Point-M2AE has been released. @adityamwagh @AnukritiSinghh
Sorry for making you wait. The fine-tuning code of Point-M2AE has been released. @AB1675
对的，因为clip的image encoder已经通过预训练和text encoder形成了很好的对应关系，所以可以直接进行zero-shot分类；如果使用pointnet等3d网络，需要一段额外的训练步骤，使得pointnet和clip的text encoder相对应，我们进行过尝试，这样会伤害网络的transfer能力，并且不再是3d数据上的zero-shot分类了。
Thanks for your interest! We are organizing the code and will release it in two weeks.
@Lature-Yang @jlqzzz @Fungungun Sorry for the delay! We have released the pre-training code. The fine-tuning code will be available in a few days.