ZipDepth: An ultra-lightweight neural network for depth estimation on mobile devices

🛠 ZipDepth: An ultra-lightweight neural network for depth estimation on mobile devices

ZipDepth has been introduced—a compact neural network for monocular depth estimation, optimized for operation on mobile devices and Edge platforms. The model contains only 6.1 million parameters and was trained using knowledge distillation from the heavy Depth Anything v2-Large.

🌍 ZipDepth demonstrates the capability of effective knowledge distillation from giant foundation models to ultra-lightweight architectures without critical loss of quality, which is crucial for mobile NPUs and DSPs.

👤 High-precision depth estimation is now available in real-time on smartphones and embedded systems (e.g., Jetson Orin), opening new possibilities for mobile AR and robotics.

Source 1: https://zipdepth.github.io/ Source 2: https://github.com/fabiotosi92/ZipDepth