search

ranahanocka / MeshCNN

Convolutional Neural Network for 3D meshes in PyTorch
MIT License
1.61k stars 320 forks source link

Error when using own dataset #79

Closed AhmadObeid closed 4 years ago

AhmadObeid commented 4 years ago

Thanks for the upload. I am facing an error when attempting to use my own data for experimentation. I made sure to place all files correctly the way you have done in the used dataset shrec_16. The error I am getting says:

Index can't contain negative values

My dataset is the shrec_17 dataset, where I have patches extracted from the different meshes, and I want to classify them. To reproduce the error, I am attaching a zip folder containing examples of these patches.

Examples.zip

Thank you in advance.

ranahanocka commented 4 years ago

Hi @AhmadObeid ,

I think you should reduce the amount of pooling (increase the--pool_res flag). Right now by default it is: --pool_res 600 450 300 180 for classification. How about --pool_res 600 450 300 210 ? If that doesn't work, what about --pool_res 600 450 300 270. You can continue to increase the resolution until you find something that works well with your dataset. Also, make sure that set--ninput_edges correctly. See the wiki for more info.

AhmadObeid commented 4 years ago

Hello, thanks for the answer. Apologies, I should have closed the issue earlier, as I have already found my solution in other closed issues. Namely, issues #60 and #18 .