2D image obscures content more than 500m from the camera #8373
Labels
Comments
jleibs
added a commit
that referenced
this issue
Dec 9, 2024
### What
This adjusts the value to now be in physical scene units. I.e. for most
3d scenes in meters, the clip plane will be 10cm, which is a very
reasonable default.
Additionally, this gives users an escape hatch for scenes where the
default clipping plane causes their data to not be visible.
Test script:
```python
import rerun as rr
import rerun.blueprint as rrb
import numpy as np
rr.init("rerun_example_pinhole_perspective", spawn=True)
img = np.zeros((300, 300, 3), dtype=np.uint8)
cam_proj = [152.34006, 0, 175.5, 0, 152.34006, 175.5, 0, 0, 1]
rr.log(
"world/cam",
rr.Pinhole(
image_from_camera=cam_proj, width=300, height=300, image_plane_distance=0.1
),
)
rr.log("world/cam", rr.Image(img))
rr.log(
"world/points",
rr.Points3D([(0.0, 0.5, 0.1), (0.1, 0.1, 0.3), (-0.1, -0.1, 0.5)], radii=0.025),
)
rr.log_file_from_path("examples/assets/example.gltf")
rr.log("examples/assets", rr.Transform3D(translation=[0, 0, 0.4], scale=0.05))
rr.send_blueprint(
rrb.Horizontal(
rrb.Spatial3DView(),
rrb.Spatial2DView(
origin="world/cam",
contents="/**",
visual_bounds=rrb.VisualBounds2D(
near_clip_plane=0.01,
),
),
)
)
```
### New issues:
- #8373
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
The way our projection math works at the moment, all content acts as if it sits in front of a giant image sitting 500m away from the camera.
This value is hard-coded in the way we set up our transforms.
There is a lot of complexity in how we set up our transform chain for pinholes at the moment and it would be nice to simplify that as well.
The text was updated successfully, but these errors were encountered: