search

ranahanocka / MeshCNN

Convolutional Neural Network for 3D meshes in PyTorch
MIT License
1.56k stars 314 forks source link

Tagging my own 3D Data for Segmentation #68

Open nitish11 opened 4 years ago

nitish11 commented 4 years ago

@ranahanocka @emepetres @bertjiazheng : I found a link in the wiki, any update on this https://github.com/ranahanocka/MeshCNN/wiki/Segmentation ?

I want to tag parts and train my own 3D segmentation model ?

ranahanocka commented 4 years ago

Hi @nitish11 ,

There is no fully automatic tool for segmenting shapes, if there was - you probably wouldn't need to use MeshCNN -> could just use that tool instead :)

You can try searching for unsupervised shape segmentation approaches. For example, one of the co-authors of MeshCNN did such a work on shape segmentation, and another work which propagates one segmented model to others structure. There might also be some semi-automatic tools for doing this.

In a quick google search, I came across this: Protrusion-oriented 3D Mesh Segmentation

nitish11 commented 4 years ago

Hi @ranahanocka, thanks for your reply, I wanted to understand how do we create dataset for training a 3D model segmentation?

ranahanocka commented 4 years ago

Did you try going through the segmentation example from the README?

HuanDinh20 commented 2 years ago

I tried through the segmentation example from the README too. And now, I want to apply it on my own dataset, but I don't know any method to create soft_labels for seseg files