Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Mesh R-CNN Voxel Head #11

Open
wants to merge 11 commits into
base: mesh_rcnn
Choose a base branch
from
Open

Mesh R-CNN Voxel Head #11

wants to merge 11 commits into from

Conversation

zghera
Copy link
Collaborator

@zghera zghera commented Oct 19, 2021

Description

Implements the voxel prediction head of the voxel branch for the re-implementation of Mesh RCNN in TensorFlow.

Type of change

Closes #5

  • A new research paper code implementation

Tests

Testing is captured in voxel_head_test.py. The unit tests include:

  • test_network_creation: Verify the output shapes of the voxel head.
  • test_serialize_deserialize: Create a network object that sets all of its config options.
  • test_gradient_pass_though: Ensure the gradients of the layer are not None.
  • test_build_from_config: Test head creation from config and factory function.

Checklist

  • I have signed the Contributor License Agreement.
  • I have read guidelines for pull request.
  • My code follows the coding guidelines.
  • I have performed a self code review of my own code.
  • I have commented my code, particularly in hard-to-understand areas.
  • I have made corresponding changes to the documentation.
  • My changes generate no new warnings.
    |--> Only warnings are from mypy that it cannot find library stubs/implementation for various modules.
  • I have added tests that prove my fix is effective or that my feature works.

zghera added 5 commits October 9, 2021 13:24
I am nearly complete with my initial rough implementation of the voxel head. The only thing I have left is the weight and bias initializations in build(). After that, I will start testing following yolo_head_test as an exemplar.
Added support to handle both ShapeNet and Pix3D as well as initialization of layer weight and biases.
See updated questions in voxel_head.py. Mypy is also failing to find the voxel_head module right now and I cannot figure out why. I will try to look into this after finishing up testing.
@zghera zghera self-assigned this Oct 19, 2021
@zghera zghera linked an issue Oct 19, 2021 that may be closed by this pull request
3 tasks
@zghera zghera marked this pull request as ready for review October 19, 2021 22:00
@zghera zghera requested review from davidliii and lenliu3 October 19, 2021 22:00
@zghera zghera changed the title Mesh rcnn voxel head Mesh R-CNN Voxel Head Oct 19, 2021
Since these boolean's will be set at the initialization of the model, it would be cheaper to do this evaluation at graph construction (with 'if-else') time as opposed to at runtime (with 'tf.cond').
Tyan3001 pushed a commit that referenced this pull request Nov 30, 2021
* added shell scipt to prepare dataset

* fixed path errors

* added 1x and 3x experiment configs

* fixed formatting
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

Implement Voxel Branch
1 participant