create extrinsic from eye point #65
Conversation
BlendParams background_color is immutable , type hint as a sequence allows setting new values in constructor.
background_color type hint
return R, t from a camera eye point and look at target.
facebook-github-bot
added
the
CLA Signed
|
@davegreenwood, as per the the guidelines in
|
pytorch3d/renderer/cameras.py
Outdated
| T = -torch.bmm(R.transpose(1, 2), eye[:, :, None])[:, :, 0] | ||
| return R, T | ||
|
|
||
|
|
||
| def look_at_view_transform( | ||
| dist, |
There was a problem hiding this comment.
Instead of creating a new function, I would suggest passing in the camera_position as an input to look_at_view_transform. Then we can check if either the spherical angles have been provided or the camera position.
There was a problem hiding this comment.
@nikhilaravi - that's a good idea - I'll make that change.
tests/test_cameras.py
Outdated
| if __name__ == "__main__": | ||
| unittest.main() |
There was a problem hiding this comment.
Please remove these two lines.
|
@davegreenwood thanks for making the updates. I added a few more comments to the tests. Please make these changes and then I think we can merge this PR. |
tests/test_cameras.py
Outdated
| R, t = look_at_view_transform(dist, elev, azim) | ||
| # Using default dist=1.0, elev=0.0, azim=0.0 | ||
| R_default, t_default = look_at_view_transform() | ||
| # R should be identity, t should be (0, 0, 1) | ||
| R_expected = torch.tensor([np.eye(3)], dtype=torch.float32) |
There was a problem hiding this comment.
For this test it is sufficient to just check the default matches the expected.
There was a problem hiding this comment.
hi, I trimmed those unnecessary lines.
|
@davegreenwood apologies for the delay, this looks good - I will merge the PR! |
There was a problem hiding this comment.
@nikhilaravi has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
tests/test_cameras.py
Outdated
| dist = math.sqrt(2) | ||
| elev = math.pi / 4 | ||
| azim = 0.0 | ||
| eye = ((0.0, 1.0, -1.0), ) |
There was a problem hiding this comment.
Now that the camera_position_from_spherical_angles function has been modified so that z is positive, you will need to change eye to eye = ((0.0, 1.0, 1.0), ) in order for the tests to pass.
There was a problem hiding this comment.
hi @nikhilaravi, I've made that required change. The test does not pass in my branch, but if I did a rebase it would, due to line 947 in cameras.py.
|
@davegreenwood can you please make the one small change requested? After that I can merge your PR. |
There was a problem hiding this comment.
@nikhilaravi has imported this pull request. If you are a Facebook employee, you can view this diff on Phabricator.
|
@nikhilaravi merged this pull request in 2480723. |
Create extrinsic parameters from eye point.
Create the rotation and translation from an eye point, look-at point and up vector.
see:
https://www.khronos.org/registry/OpenGL-Refpages/gl2.1/xhtml/gluLookAt.xml
It is arguably easier to initialise a camera position as a point in the world rather than an angle.