spatial-maker/research/ at main · xcc.es/spatial-maker · Tangled

xcc.es / spatial-maker

0

[READ-ONLY] Mirror of https://github.com/mrgnw/spatial-maker.

0

spatial-maker / research

4 files

at main

Morgan Williams Add Rust depth estimation library with CoreML Swift bridge 5mo ago

coreml-benchmark-results.md

Add Rust depth estimation library with CoreML Swift bridge Implements a Rust library for converting 2D images to stereoscopic 3D using AI depth estimation, with two backends: - ONNX Runtime (cross-platform, 430ms/frame on M4 Pro) - CoreML via Swift bridge (macOS, 128ms/frame on M4 Pro, 3.4x faster) Includes: - Depth Anything V2 inference (Small/Base/Large models) - DIBR stereo pair generation - SBS image output - CoreML model conversion scripts (PyTorch → CoreML) - Model download script (from huggingface.co/mrgnw/depth-anything-v2-coreml) - Comprehensive documentation and benchmarks

5 months ago

coreml-conversion-results.md

Add Rust depth estimation library with CoreML Swift bridge Implements a Rust library for converting 2D images to stereoscopic 3D using AI depth estimation, with two backends: - ONNX Runtime (cross-platform, 430ms/frame on M4 Pro) - CoreML via Swift bridge (macOS, 128ms/frame on M4 Pro, 3.4x faster) Includes: - Depth Anything V2 inference (Small/Base/Large models) - DIBR stereo pair generation - SBS image output - CoreML model conversion scripts (PyTorch → CoreML) - Model download script (from huggingface.co/mrgnw/depth-anything-v2-coreml) - Comprehensive documentation and benchmarks

5 months ago

Add Rust depth estimation library with CoreML Swift bridge Implements a Rust library for converting 2D images to stereoscopic 3D using AI depth estimation, with two backends: - ONNX Runtime (cross-platform, 430ms/frame on M4 Pro) - CoreML via Swift bridge (macOS, 128ms/frame on M4 Pro, 3.4x faster) Includes: - Depth Anything V2 inference (Small/Base/Large models) - DIBR stereo pair generation - SBS image output - CoreML model conversion scripts (PyTorch → CoreML) - Model download script (from huggingface.co/mrgnw/depth-anything-v2-coreml) - Comprehensive documentation and benchmarks

5 months ago

session-summary.md

Add Rust depth estimation library with CoreML Swift bridge Implements a Rust library for converting 2D images to stereoscopic 3D using AI depth estimation, with two backends: - ONNX Runtime (cross-platform, 430ms/frame on M4 Pro) - CoreML via Swift bridge (macOS, 128ms/frame on M4 Pro, 3.4x faster) Includes: - Depth Anything V2 inference (Small/Base/Large models) - DIBR stereo pair generation - SBS image output - CoreML model conversion scripts (PyTorch → CoreML) - Model download script (from huggingface.co/mrgnw/depth-anything-v2-coreml) - Comprehensive documentation and benchmarks

5 months ago