Open3d: Allow configuration of view parameters from python

[adamlofts]

I would like to directly set the view parameters rather than call rotate(), translate() ...

From C++ it looks like I can use this function:

https://github.com/IntelVCL/Open3D/blob/995b646375121e348ff814b138a924bfac0f047e/src/Visualization/Visualizer/ViewControl.cpp#L96

Can you add this call to python?

[syncle]

Thanks for the request, but I am occupied by some other issues. I will handle this issue soon.

[syncle]

@adamlofts: I am working on this.

Saving viewpoint

Actually direct saving/loading view point is done by using pinhole_camera_trajectory. For example:

    vis = Visualizer()
    vis.create_window()
    ctr = vis.get_view_control()
    pcd = read_point_cloud("../../TestData/fragment.pcd")
    vis.add_geometry(pcd)
    vis.run() # user changes the view and press "q" to terminate
    param = ctr.convert_to_pinhole_camera_parameters()
    trajectory = PinholeCameraTrajectory()
    trajectory.intrinsic = param[0]
    trajectory.extrinsic = Matrix4dVector([param[1]])
    write_pinhole_camera_trajectory("test.json", trajectory)
    vis.destroy_window()

will save test.json that has single camera pose in camera trajectory format.

Loading viewpoint

    vis = Visualizer()
    vis.create_window()
    ctr = vis.get_view_control()
    pcd = read_point_cloud("../../TestData/fragment.pcd")
    trajectory = read_pinhole_camera_trajectory("test.json")
    vis.add_geometry(pcd)
    ctr.convert_from_pinhole_camera_parameters(
            trajectory.intrinsic, trajectory.extrinsic[0])
    vis.run()
    vis.destroy_window()

will read test.json and display the same view.

However, as @adamlofts indirectly pointed out, directly working with pinhole_camera_paramters is non-trivial as there is no I/O for pinhole_camera_paramters.

@qianyizh: do you hide I/O functions for PinholeCameraParameters intentionally, or it is not exposed?

[qianyizh]

It is a IJsonConvertible, so the IO should be exposed by default?

[syncle]

@qianyizh: can you tell me the name of the function? I think we need to expose PinholeCameraParameters first, right?

Here are exposed classes for your reference:

    open3d - Python binding of Open3D

CLASSES
    pybind11_builtins.pybind11_object(builtins.object)
        ColorMapOptmizationOption
        CorrespondenceChecker
            CorrespondenceCheckerBasedOnDistance
            CorrespondenceCheckerBasedOnEdgeLength
            CorrespondenceCheckerBasedOnNormal
        DoubleVector
        FastGlobalRegistrationOption
        Feature
        Geometry
            Geometry2D
                Image
            Geometry3D
                LineSet
                PointCloud
                TriangleMesh
        GlobalOptimizationConvergenceCriteria
        GlobalOptimizationMethod
            GlobalOptimizationGaussNewton
            GlobalOptimizationLevenbergMarquardt
        GlobalOptimizationOption
        ICPConvergenceCriteria
        ImageFilterType
        IntVector
        KDTreeFlann
        KDTreeSearchParam
            KDTreeSearchParamHybrid
            KDTreeSearchParamKNN
            KDTreeSearchParamRadius
        Matrix4dVector
        OdometryOption
        PinholeCameraIntrinsic
        PinholeCameraIntrinsicParameters
        PinholeCameraTrajectory
        PoseGraph
        PoseGraphEdge
        PoseGraphEdgeVector
        PoseGraphNode
        PoseGraphNodeVector
        RANSACConvergenceCriteria
        RGBDImage
        RGBDOdometryJacobian
            RGBDOdometryJacobianFromColorTerm
            RGBDOdometryJacobianFromHybridTerm
        RegistrationResult
        RenderOption
        SelectionPolygonVolume
        TSDFVolume
            ScalableTSDFVolume
            UniformTSDFVolume
        TSDFVolumeColorType
        TransformationEstimation
            TransformationEstimationPointToPlane
            TransformationEstimationPointToPoint
        Vector2iVector
        Vector3dVector
        Vector3iVector
        VerbosityLevel
        ViewControl
        Visualizer
            VisualizerWithEditing
            VisualizerWithKeyCallback

[qianyizh]

I think we have changed it to PinholeCameraIntrinsic in a previous version? And it is exposed in python interface.

[sasa2124]

Hello everyone, I would like to capture depth image from 3d point and I want to save viewpoint from different views however I got some problem with respect to PinholeCameraTrajectory. Here is my code

from open3d import *
import numpy
import transforms3d
from scipy.spatial.transform import *
import matplotlib.pyplot as plt

pcd = open3d.io.read_point_cloud("scene0173.xyz")

vis = Visualizer()
vis.create_window()
ctr = vis.get_view_control()
vis.add_geometry(pcd)
vis.run() # user changes the view and press "q" to terminate
param = ctr.convert_to_pinhole_camera_parameters()
trajectory = open3d.camera.PinholeCameraTrajectory()
trajectory.intrinsic = param[0]

TypeError: 'open3d.open3d.camera.PinholeCameraParameters' object does not support indexing

Please help me with this problem