who can tell me how to deal this problem?

[wuyouqian3]

trafficstars

loading dataset creating model start training Traceback (most recent call last): File "train_AVD.py", line 77, in loss = model(obs_batch, valid_pt, pose_batch) File "/home/thebs/anaconda3/envs/fmr/lib/python3.6/site-packages/torch/nn/modules/module.py", line 477, in call result = self.forward(*input, **kwargs) File "../models/deepmapping.py", line 114, in forward self.obs_local, self.n_samples,sensor_center) File "../models/deepmapping.py", line 35, in sample_unoccupied_point unoccupied[:, (idx - 1) * L:idx * L, :] = center + (local_point_cloud-center) * fac RuntimeError: The size of tensor a (3) must match the size of tensor b (2) at non-singleton dimension 2

[Puland502]

Did you solve this problem? how is it solved

[stevenshinechen]

I resolved the error by changing sensor_center = sensor_pose[:,:,:2] to sensor_center = sensor_pose[:,:,:3] in DeepMapping_AVD This matches the dimensions of unoccupied[:, (idx - 1) * L:idx * L, :] and center.

[DineshchandarR]

After using this method to solve the issue, does the model converge and generate good results for AVD?

[dmax123]

If you did not use warm start on the AVD dataset (the default setting in train_AVD.py), then @Basekill's approach should solve the issue. No warm start was applied to AVD dataset because the overlap between two successive frames is very limited. Warm start using ICP did not perform well.

[simbaforrest]

@wuyouqian3 and others, if @dmax123's answer addressed your questions, please consider closing this issue. Meanwhile, I would suggest you take a look at our newest CVPR2023 work along this line but for larger-scale LiDAR point cloud mapping that DeepMapping1 cannot handle: https://github.com/ai4ce/DeepMapping2

[simbaforrest]

It seems that this thread is idle. So I am closing it now. Feel free to reopen it if anyone has additional questions.