search

autonomousvision / convolutional_occupancy_networks

[ECCV'20] Convolutional Occupancy Networks
https://pengsongyou.github.io/conv_onet
MIT License
830 stars 113 forks source link

How can I generate mesh with my own point cloud data #54

Closed zhao-you-fei closed 1 year ago

zhao-you-fei commented 1 year ago

@nynyg @pengsongyou @haofeixu How can I generate mesh with my own point cloud data, and what are the other input data? Looking forward to your reply!

Barnonewdm commented 1 year ago

It could be very helpful if the pipeline of preparation of the training dataset could be supplied.

pengsongyou commented 1 year ago

@nynyg @pengsongyou @haofeixu How can I generate mesh with my own point cloud data, and what are the other input data? Looking forward to your reply!

Hi @zhao-you-fei sorry for the delay. If you want to reconstruct from your own point cloud, please check how we organize our data in the provided demo data. You can download the data by bash scripts/download_demo_data.sh, and read different data fields in numpy array.

pengsongyou commented 1 year ago

It could be very helpful if the pipeline of preparation of the training dataset could be supplied.

@Barnonewdm for shapenet data, ONet provided how they prepare the training set here.

As for the synthetic indoor dataset, you can find how we prepare in the last paragraph of this.

Barnonewdm commented 1 year ago

It could be very helpful if the pipeline of preparation of the training dataset could be supplied.

@Barnonewdm for shapenet data, ONet provided how they prepare the training set here.

As for the synthetic indoor dataset, you can find how we prepare in the last paragraph of this.

Thanks a lot! A quick question. Do we need to prepare mesh as the ground truth, or just points should be enough?

pengsongyou commented 1 year ago

@Barnonewdm Check the demo data, you do not need to have GT mesh. Points should be fine.

Barnonewdm commented 1 year ago

@Barnonewdm Check the demo data, you do not need to have GT mesh. Points should be fine.

I mean for the training. Only points are needed for training the model?

pengsongyou commented 1 year ago

We need ground truth occupancy for training, so you need GT mesh to create occupancy.

On Wed, 14 Dec 2022, 06:57 Barnone @.***> wrote:

@Barnonewdm https://github.com/Barnonewdm Check the demo data, you do not need to have GT mesh. Points should be fine.

I mean for the training. Only points are needed for training the model?

— Reply to this email directly, view it on GitHub https://github.com/autonomousvision/convolutional_occupancy_networks/issues/54#issuecomment-1350440513, or unsubscribe https://github.com/notifications/unsubscribe-auth/ADNJ3IXKW3POMZVDFBHI6UTWNFOWHANCNFSM6AAAAAARM4LDT4 . You are receiving this because you modified the open/close state.Message ID: <autonomousvision/convolutional_occupancy_networks/issues/54/1350440513@ github.com>