This simple proof of concept uses the CoreML Depth Anything v2 Small model to convert an image into a 3D point cloud file.
It’s ideal for mapping images into point clouds for 3D reconstruction or analysis.
Note: You may need to adjust the aspect ratio of input images to get accurate results.
Pass what ever image as file argument. Gives you the depth map of the given image
Pass what ever image as file argument. Gives you the point cloud file (data) for the given image
This model being used Depth Anything V2 Core ML Models