From OrbbecSDK to OrbbecSDK_v2, basic differences.

[bumbastic]

Switching an App's OrbbecSDK dependencies to OrbbecSDK_v2. Results i heavy distorted/wrong depth images.

I believe this is the only code where I handle the depth image outside the SDK:

    auto depth_frame = frame->as<ob::VideoFrame>();
    const uint32_t width = depth_frame->width();
    const uint32_t height = depth_frame->height();

    uint16_t* src_data = (uint16_t*)(void*)depth_frame->data();

    // Create an open3d::t::geometry::Image from camera frame data
    open3d::core::Tensor tensor(src_data, open3d::core::Dtype::UInt16, { height, width, 1 }, {});
    open3d::t::geometry::Image open3d_image(tensor);

visualizing open3d_image as 3d point cloud, shows a heavily distorted depth image. What changed?

I expected same result for both versions of the SDK.

• OS: Windows 11.
• SDK: 2.0.18 vs 1.10.16
• Model: Femto MEGA.
• Firmware Version: 1.3.0.

[zhonghong322]

After enabling hardware D2C in OrbbecSDK v2, the depth point cloud had distortion, which has been fixed in version v2.1.1.
https://github.com/orbbec/OrbbecSDK_v2/releases/tag/v2.1.1