Code Options

All the general options are defined in base_options.py. Where as train specific options are in train_options.py and inference (test) options are in test_options.py

Running on CPU

set --gpu_ids -1

Input Mesh Resolution

--ninput_edges should be the maximum number of edges in the entire dataset. Any meshes which have less than that will be padded automatically.

Pooling Resolution

This is a hyper parameter of the network. For example setting --pool_res 600 450 will create two mesh_pooling layers in the network, where the first one pools meshes to 600 edges, and the second one pools to 450 edges. If using meshes with larger resolutions than the examples from this repo, be sure to increase the pooling resolution.