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

Init point cloud and video augmentation module #2374

Closed
wants to merge 6 commits into from

Conversation

shijianjian
Copy link
Member

Changes

Fixes # (issue)

Type of change

  • 📚 Documentation Update
  • 🧪 Tests Cases
  • 🐞 Bug fix (non-breaking change which fixes an issue)
  • 🔬 New feature (non-breaking change which adds functionality)
  • 🚨 Breaking change (fix or feature that would cause existing functionality to not work as expected)
  • 📝 This change requires a documentation update

Checklist

  • My code follows the style guidelines of this project
  • I have performed a self-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
  • Did you update CHANGELOG in case of a major change?

@shijianjian shijianjian marked this pull request as draft May 10, 2023 20:31
Copy link
Member

@edgarriba edgarriba left a comment

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@shijianjian shouldnt we define a Pointcloud data structure too ? Do you have something in mind whether to represent dense or sparse pointcloud and how to deal with the memory footprint for the dense case ?

Comment on lines 42 to 62
def transform_tensor(self, input: Tensor) -> Tuple[Tensor, Optional[Tensor], Optional[Tensor], Optional[Tensor]]:
"""Split a given tensor to [xyz, normal, rgb, others].

Of which xyz is required, while the rest are optional.
"""
C = input.size(-1)
if C < 3:
raise RuntimeError(
f"Insufficient input channels. We expect at least (B, N, C) and C >= 3. Got {input.shape}."
)
if C == 3:
return input, None, None, None
if C < 6:
return input[..., :3], None, None, input[..., 3:]
if C == 6:
return input[..., :3], input[..., 3:], None, None
if C < 9:
return input[..., :3], input[..., 3:6], None, input[..., 6:]
if C == 9:
return input[..., :3], input[..., 3:6], input[..., 6:], None
return input[..., :3], input[..., 3:6], input[..., 6:9], input[9:]
Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

@edgarriba The fields for a point cloud object can be represented like above.

Also, I think we should support .from_open3d and .to_opend3d.
http://www.open3d.org/docs/release/python_api/open3d.geometry.PointCloud.html

Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

100% agree

@cjpurackal
Copy link
Member

@shijianjian shouldnt we define a Pointcloud data structure too ? Do you have something in mind whether to represent dense or sparse pointcloud and how to deal with the memory footprint for the dense case ?

This sounds good. I've been trying out some RGBD stuff and was wondering the same. Perhaps we can open a different discussion thread(issue) for this.

@shijianjian
Copy link
Member Author

@shijianjian shouldnt we define a Pointcloud data structure too ? Do you have something in mind whether to represent dense or sparse pointcloud and how to deal with the memory footprint for the dense case ?

This sounds good. I've been trying out some RGBD stuff and was wondering the same. Perhaps we can open a different discussion thread(issue) for this.

@edgarriba @cjpurackal Can you point me some links as references? I am not sure what to do on this.

Probably worth having RGBD data structure.

@edgarriba
Copy link
Member

@cjpurackal can you start that thread ? Very often RGBD are just 4xHxW but depending how we want to target pointclouds it can be defined as a just as sparse vectors

@cjpurackal
Copy link
Member

@cjpurackal can you start that thread ? Very often RGBD are just 4xHxW but depending how we want to target pointclouds it can be defined as a just as sparse vectors

will do

@johnnv1
Copy link
Member

johnnv1 commented Mar 14, 2024

as part of #2719 plan, we have deleted the master in favor of main, if anyone has an interest in keep/restart working on this, let me know to help you.

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

Successfully merging this pull request may close these issues.

None yet

4 participants