…els into mesh_rcnn_lintergit

Added first implementation for graphconv. The code was validated with a simple test on an undirected graph.

Implementation of graph conv with comprehensive testing completed.

…ed/padded representations of vertices and faces and implemented edge computation

…t run slower, but will be easier to debug in the future

Added gradient test and used pylint to fix formatting issues

This was done so the mesh sampler and the utility functions/class for getting the closest points in each point cloud can both live in mesh sample while providing a clear distinction between them.

Note: this is currently a prototype. A rationale for pushing this prototyping and the TODOs are listed in the module docstring.

Added class that encompases all of the mesh based loss terms (chamfer, normals, and edge) as well as voxel BCE loss. Added option to pass batch weights to chamfer and normal loss. Refactored some code for greater modularity.

TF implementation was applying point reduction before multiplying by weights when the original implementation was doing the opposite.

Edge loss worked after merging in the cubify fixed from mesh_rcnn_mesh_refinement branch (commits 2c516d4, 0aabf68, e53189f, 8eb0106, 2ea6446).

These tests now run faster and no longer assert values for functions that depend on other random functions (i.e. the mesh sampler).

Mesh R-CNN Mesh Losses

Fully unit tested (test_z_head.py)
Weights loaded (test_load_weights.py)
Differential tested (code not provided)

added config in mesh_rcnn.py
Fixed fc layer weight initializer

newlines at the end of files

Aligning function arguments to be in same column

test_z_head main function now says "tf.test.main()"

Add parameters to z-head test

Mesh rcnn zhead

…o mesh_rcnn