The current API doesnot supports PointCloud for open3d.pipelines.integration.TSDFVolume.integrate.
Usually, we have the Extrinsics of cameras before the Volumetric integration, so there is no need to do the depth to Pcd generation again, which is just time-consumming duplicate mostly.
Reading the C++ code, it only needs to add one interface function. Please add it in the next version.