How do I run inference on one of the model

Julie-tang00 / Point-BERT

[CVPR 2022] Pre-Training 3D Point Cloud Transformers with Masked Point Modeling

MIT License

542 stars 65 forks source link

How do I run inference on one of the model #68

Closed m-a-x-c closed 1 year ago

m-a-x-c commented 1 year ago

I am an inexperienced user. I wan to run inference on a dataset to predict its shapes.

what has to be the file format of the point cloud files (e.g. .obj)?
how do I structure the directory of my files?
where can more information on this, as I don't see such information in the README?

m-a-x-c commented 1 year ago

So having looked at datasets/ModelNetDataset.py it seems that for each point cloud file, I have to give a file with comma separated values for each component of the vertex (e.g. V1_X, V1_Y, V1_Z, V1_NX, V1_NY, V1_NZ, V2_X, V2_Y, V2_Z, V2_NX, V2_NY, V2_NZ, ...). Is that correct?

Is this a common file format in the point cloud world? Some reference material regarding this would be nice.

m-a-x-c commented 1 year ago

I think I have finally figured out how to use a custom dataset. I have to write a python file in datasets/DatasetNameDataset.py which includes a class named DatasetName and the functions init(), getitem(), and len(). The actually files can be in any format, but I have to transform it into the specific numpy array format.