kevinqz/Depth-Anything-V2-Small-CoreAI

Canonical: kevinqz/Depth-Anything-V2-Small-CoreAI — source of truth.

Depth Anything V2 Small (fabric)

An Apple Core AI conversion of depth-anything/Depth-Anything-V2-Small-hf — a monocular depth estimator that maps a single RGB image to a per-pixel (inverse-)depth map. Produced by coreai-fabric and indexed by coreai-catalog.

Dense prediction, fully on-device. This is a single-forward dense predictor — RGB image (1×3×518×518) in, per-pixel depth (1×518×518) out, no loop. The host owns only image preprocessing (resize to a multiple of 14, ImageNet normalize) and any depth colormap for display.

Model facts

Field	Value
Parameters	0.025B
Architecture	cnn/transformer
Capabilities	monocular-depth
Input	1×3×518×518
Output (depth)	1×518×518
Quantization / precision	none / float32
On-disk size	94 MB
Asset kind	single-graph dense predictor (image -> per-pixel depth)
assetVersion	2.0

Use it — this needs host code you supply

The bundle is a single static-size graph: pixel_values (1×3×518×518) in → predicted_depth (1×518×518) out. You supply the resize-to-multiple-of-14 + ImageNet normalization and any visualization/colormap in your host code (Swift or Python).

pip install coreai-catalog && coreai-catalog install depth-anything-v2-small

Requirements

• Deployment: macOS 27.0+ / iOS 27.0+, Xcode 27+. The asset serializes with minimum_os v27, so the on-device Swift runtime requires macOS/iOS 27+. A Mac on macOS 26 can convert and inspect it but not run it on-device.
• Apple Silicon.

Verification (output parity)

• Gate A (structure): passed — the bundle's layout + metadata were validated; the graph loads.
• Gate B — graph_output_cosine: 1.000000 min output cosine (median 1.000000) vs the fp32 torch depth model over 8 seeded pixel_values, measured on apple_silicon. Certifies the export computes the SAME output as the source — a conversion-fidelity metric, not task accuracy.
• This certifies the export is numerically faithful to the source model — it does NOT certify absolute depth accuracy on your images. Reproduce with coreai-fabric verify.

Provenance

Field	Value
Base model	depth-anything/Depth-Anything-V2-Small-hf @ 5426e4f0f36572d16453bbda7a8389317b1bef99
Converted by	models/depth_anything/export.py (version not reported)
Recipe	depth-anything-v2-small (recipe_source: fabric)
Precision / quantization	float32 / none
Conversion date	2026-07-10

Machine-readable, in this repo: parity-report.json · reproduce-manifest.json · LICENSE.

License and attribution

Weights licensed apache-2.0 — see the bundled LICENSE. This artifact is a converted derivative of the base model: its weights were converted to Apple Core AI format. The conversion itself is community work.

Links

• Base model: depth-anything/Depth-Anything-V2-Small-hf
• Reproduce: recipe depth-anything-v2-small
• Index: coreai-catalog
• HF Collection

The on-device Core AI ecosystem

• coreai-fabric — the reproducible recipe → .aimodel pipeline that produced this asset.
• coreai-catalog — the index of Core AI models with provenance and integration snippets.
• apple/coreai-models — Apple's official exporters and runtimes.

Not affiliated with Apple

Community conversion. Not produced, hosted, or endorsed by Apple. Apple and Core AI are trademarks of Apple Inc., used here only to describe the target runtime/format.