KeighBee/coreml-DepthPro

DepthPro CoreML Models

DepthPro is a monocular depth estimation model. This means that it is trained to predict depth on a single image.

DepthPro paper

DepthPro original repo

Model Variants

Variant	Size (MB)
DepthPro	1900
DepthPro: Pruned 10% Sparsity, Quantized Linear Symmetric	1100
DepthPro Normalized Inverse Depth	1290
DepthPro Normalized Inverse Depth: Pruned 10% Sparsity, Quantized Linear Symmetric	745

Model Inputs and Outputs

DepthPro Normalized Inverse Depth Models

Inputs

• pixel_values: 1536x1536 3 color image.

Outputs

• normalized_inverse_depth 1536x1536 monochrome image.

DepthPro Models

Inputs

• pixel_values: 1536x1536 3 color image.
• original_widths: 1x1x1x1 Tensor containing the original width of the image before resizing.

Outputs

• depth_meters: 1x1x1536x1536 Tensor containing depth in meters.

Download

Install huggingface-cli

brew install huggingface-cli

To download one of the .mlpackage folders to the models directory:

huggingface-cli download \
  --local-dir models --local-dir-use-symlinks False \
  KeighBee/coreml-DepthPro \
  --include "DepthProNormalizedInverseDepth-pruned10-Qlinear.mlpackage/*" "DepthPro-pruned10-Qlinear.mlpackage/*"

To download everything, skip the --include argument.

Integrate in Swift apps

The huggingface/coreml-examples repository contains sample Swift code for DepthProNormalizedInverseDepth-pruned10-Qlinear.mlpackage and other models. See the instructions there to build the demo app, which shows how to use the model in your own Swift apps.