Xiaopei Wu
Xiaopei Wu
Hi, it is a bug in SFD code when there is no pseudo point in any ROI. Sometimes, we just resume the latest checkpoint and the error will disappear.
Hi, we didn't encounter your problem that the training doesn't start at all, maybe you can skip forward and backward when this situation happened.
Hi, sorry for the late reply, we did not train the TWISE in our dataset. We directly use the checkpoint TWISE provide to generate pseudo point clouds.
Hi, sorry for the late reply, could you change the batch size of each GPU to 1? And do you use spconv2x? Our code is based on spconv1x. You can...
Hi, sorry for the late reply, you can modify our yaml file to support 3 classes according to [pv_rcnn.yaml](https://github.com/open-mmlab/OpenPCDet/blob/master/tools/cfgs/kitti_models/pv_rcnn.yaml). We did not do special design for ped and cyc when...
是的，您不需要修改
Hi, you need to ``` cd pcdet/ops/iou3d/cuda_op python setup.py install ```
Hi, sorry for the late reply, the released model is for cars only and there is no plan to release the model for ped and cyc. You can modify our...
Hi, sorry for the late reply, we did not calculate the number of trainable parameters of SFD. SFD consists of the detector model and depth completion model, you can refer...
Hi, sorry for the late reply. It seems that the 'self.voxel_generator' in your code is None. You need to check your spconv and do you use spconv2.x ? The code...