how to get all the points of point cloud?

[minwang-ai]

Hi Yisheng,

You transformed each pixel on depth images to the same amount of points in the point cloud with camera intrinsic matrix in this way. How did you (FFB6D and PVNet3D as well as Densefusion) get the points of occlusion parts? Did you use 3D model matching?

Best, Min

[ethnhe]

Do you mean the occluded parts that are invisible from the camera? No, we didn't use 3D model matching. We only use the visible point cloud from the camera.

[minwang-ai]

你的意思是相机看不见的被遮挡部分？不，我们没有使用 3D 模型匹配。我们只使用来自相机的可见点云。

Thank

Do you mean the occluded parts that are invisible from the camera? No, we didn't use 3D model matching. We only use the visible point cloud from the camera.

Yes. Thank you for your reply! Do you mean only use 3D models of objects for render using raster_triangle?

[ethnhe]

We use the scene point cloud visible from the camera as inputs to FFB6D. For raster_triangle, we use the reconstructed object meshes from the dataset creators for synthesis scene rendering.