mindqtrl/DA3-LARGE-1.1-GGUF-STRIPPED

This repository contains custom GGUF conversions of the original Depth-Anything-3 model, limited to monocular depth map creation.

Optimized Variant (Stripped): This model has been surgically optimized to remove layers not needed for monocular depth map creation (pose estimation, auxiliary heads, etc.). This reduces file size by up to 74% and lowers memory usage and latency.

Original Model

Modification Notice

This model is a derivative work. The weights have been converted to GGUF and have had non-essential layers removed by the MindQTRL team to optimize for single-image depth estimation.

Format details

This repository contains the stripped GGUF converted weights.

Downloads last month
7
GGUF
Model size
0.3B params
Architecture
depth-anything
Hardware compatibility
Log In to add your hardware

4-bit

5-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support