In this function we are going to evaluate the instructions in a topological walk in the way that one would in a CAD system.

Extract the coedge features from each face of the body

The coedge LCS is a special coordinate system which aligns with the B-Rep geometry. - The origin will be a

The coedge LCS is a special coordinate system which aligns with the B-Rep geometry. - The origin will be a

Extract a coedge grid (aligned with the coedge direction). The coedge grids will be of size [ num_coedges x 12 x n

Extract coedge grids (aligned with the coedge direction). The coedge grids will be of size [ num_coedges x 12 x nu

The flags for each coedge telling us if it is reversed wrt its parent edge. Notice that when coedge features are create

Extract the edge features from each edge of the body

Extract the face features from each face of the body

Extract a UV-Net point grid from the given face. Returns a tensor [ 7 x num_pts_u x num_pts_v ] For each point the val

Extract a UV-Net point grid for each face. Returns a tensor [ num_faces x 7 x num_pts_u x num_pts_v ] For each point t

The scale factors which need to be applied to the LCS for scale invariance

Try to locate the dir where the labels are stored if this is not given explicitely

Find the loss given the logits and labels

Each edge in the B-Rep has two coedges. In this function we perform an element-wise max pooling of the feature vectors for the two co

Find the number of coedges on the given face

Find the number of coedges on each face

Find the number of faces in each brep

Find the predicted classes from the un-normalized segmentation scores

Try to find the segment names file in some usual places

Get the box of the face. We do this by using the triangles as other methods appear to give boxes which we too big. The bo

Create a pathname for the cache file for the batch with batch_idx

Create a pathname for the cache file for the batch with idx

Get the triangles from this face

Get the pathname of the file which gives the face index of each triangle in the mesh

Get the index of the edge

Get the index of the face to which this coedge belongs

Get the index of the mating coedge

Get the index of the next coedge

Get the index of the previous coedge

Get the pathname of the OBJ file for the Fusion mesh

We want to be sure that the cache is correctly built given all the hyper-parameters of the network. Here we make a hash of

We want to be sure that the cache is correctly built given all the hyper-parameters of the network. Here we make a hash of

Create a hash of the strings in the list

Create a hash of the strings in the list

Load the batch with batch_idx from the raw json data, then save a cache of the binary tensors

Load the body with idx from the STEP data, then save a cache of the binary tensors

Load data from the json files and generate a binary cache of this information

Load the data for the batch and apply the normalization

Load cache data for the batch with the given pathname

Load the data for a body.

Load cache data for the body with the given pathname

Load the body from the step file. We expect only one body in each file

Load face label file

Load feature file

Load a list of parts from a STEP file and also return a map from the hash value of the shape to the Fusion face index

Load the body from the step file. We expect only one body in each file

Load the topology file

Save the cache data for the batch

Save the face embeddings for this batch