Creating a learning model (machine learning) from 3D models

Question

I have a database with almost 20 kilogram 3D files, these are drawings from machine parts developed in CAD software (solid work). I'm trying to create a trained model from all these three-dimensional models, so I can create a 3D object recognition application when someone can take a picture from one of these parts (in the real world), and the application can provide useful information about the material, size, treatment and so on.

If someone already does something similar, any information that you can provide to me will be very grateful!

Answer 1

Some ideas:

1) Several shots : instead of one . As Rodrigo commented and Brad Larson tried to get around his method, the problem with the user accepting only one drawing for input is that you do not have enough information to create a triangulation and form a point cloud in 3D. With 4 shots taken from a slightly different angle, you can already restore parts of the subject. Comparison of point clouds will facilitate the operation of any ML algorithm, neural networks (NN), vector machine support (SVM), or others. A common standard for creating point clouds is ASTM E2807, which uses the e57 file format.

.

2) . , , , . , " ", , , , . , . CAD-, , "", .

1) , .

3) . . , . . ( " , []?" ). .

Answer 2

- , , , .

, 3D-, , . . , , Voxnet. , .

Answer 3

Pre-Trained 3D Deep Neural Networks, , .