felipebudinich/depth-map-exporter

depth-exporter is a command line tool for Mac computers with Apple Silicon chips. It takes a regular video file in MP4 format and creates a new video that shows depth information as a grayscale image, where lighter or darker areas represent how near or far things are from the camera. It does this using Apple's own tools, AVFoundation and Core ML, so it does not need FFmpeg, Python, Node.js, Electron, or other common video and AI tools installed on your machine. To use it, you need a Mac running macOS 13 or newer, the Xcode command line tools, and a Core ML depth model file that you download separately, such as the Depth Anything V2 Small model mentioned in the documentation. You build the tool yourself using Swift, then run it from the terminal with your input video and chosen model. The tool offers many optional settings. You can choose which hardware runs the AI model, such as the CPU, GPU, or Neural Engine, pick how the output looks in grayscale or inverted grayscale, and control how the depth values are normalized across frames so brightness stays consistent instead of flickering. You can also cap the output resolution, set the video bitrate, keep the original audio track, or create a side by side comparison video showing the original footage stacked above its depth map. Besides depth mapping, the project includes a second command called track-boxes. This draws colored boxes around people in a video and follows them as they move, using Apple's Vision framework for detection instead of the depth model. You can set how many people to track, provide starting box positions manually, and export debug data describing each tracked box over time. This is a specialized tool built for people comfortable with the command line and building software from source code on a Mac. It has no graphical interface, so someone unfamiliar with terminal commands or compiling Swift projects may find it hard to use directly. There is no license information provided in the material available.