Improve depth conversion precision, use tighter clamp bounds, enable double-precision arithmetic, enhance quaternion normalization, adjust validation tolerances, add per-output and angular error checks, refine diagnostic logging.

Fix quaternion handling in model output, add depth/position stats logging, update validation logic, normalize quaternions, improve rotation accuracy checks

Improve quaternion normalization stability, handle large model ONNX exports, fix CoreML framework mapping, enhance quaternion validation checks.

Add ONNX support with conditional checks, implement safe clamp/division modules, enhance conversion logic via pre-warmed ONNX export, add dynamic fallback to direct tracing, standardize output naming and metadata.

Fix typo in model conversion, enable depth alignment, improve tracing stability, add deterministic seeds, update output naming, refine descriptions, adjust validation tolerance, log debug info, set default precision to float32

Add Core ML converter script, remove prompt scripts, enable preprocessing validation, support FLOAT16/FLOAT32 precision.