Releases: google-research/visu3d
Releases · google-research/visu3d
v1.5.3
v1.5.2
- Fix import error in Python 3.11
- Now require Python 3.9 (drop 3.8 support)
v1.5.1
- Add
torch==2.0.0
support
v1.5.0
v1.4.0
- DataclassArrays:
- Added:
v3d.Spec360
camera spec to support 360 videos.
- Added:
v3d.math
:- Added:
v3d.math.interp_img
- Added:
v3d.math.carthesian_to_spherical
and
v3d.math.spherical_to_carthesian
- Added:
- Visualization:
- Added:
v3d.make_fig
supportsv3d.make_fig(a, b, c)
(in addition of
v3d.make_fig([a, b, c])
) - Added:
v3d.fig_config.cam_scale
to globally control the size of the
cameras. - Changed: Do not add
0
suffix in the figure labels (for the first
element) - Fixed:
v3d.make_fig(pts, num_samples_point3d=None)
works (previously
the argument had no effect).
- Added:
- Fixed: Subsampling for displaying simple
np.array
point clouds (regression
from previous release).
v1.3.0
Added
v3d.math
to expose:- Rotation utils (convert from/to rotation matrix)
- Subsampling util (
v3d.math.subsample
) - Interpolation util (
v3d.math.interp_points
)
DataclassArray
now supports dynamic shape fields (shape withNone
), like
Array[..., None, None]
.- More trace customization options:
- All
v3d.DataclassParams
objects now have a.fig_config
property
(customizable withobj = obj.replace_fig_config(**options)
). - Automatic subsample customizable with
points.replace_fig_config(num_samples=10_000)
. - Object names displayed in plotly are customizable using
points.replace_fig_config(name='My point cloud')
. v3d.fig_config.num_sample_xyz = 123
to overwrite the default number of
sampled rays, points,... (None
for no subsampling)
- All
DataclassArray.__dca_params__
can be set tov3d.DataclassParams
to
configure the dataclass options.
Changed
v3d
dataclass array implementation has been moved to it's independent
dataclass_array
package.- Any object implementing the
.make_traces
protocol is not visualizable by
v3d.make_fig
. No need to inheritv3d.Visualizable
anymore. - Support more complex
DType
(FloatArray
,... acceptbfloat16
,
float64
,...).
Fixed
- Translation by subtraction (
ray - np.array([0, 0, 0])
) now works FigConfig
property (cam.fig_config
,...) works correctly for batched
camera.
v1.2.0
Changed
- Camera are now displayed with a complete frame.
v1.0.0: Release version 1.0.0
Release of visu3d \o/