High-performance host-side stereo processing for deterministic, geometry-first 3D reconstruction.
CoreStereo is a host-side stereo processing module within the EdgeTrack architecture. It is responsible for converting synchronized stereo image streams into disparity, depth, and local 3D representations.
The system is designed around a RAW-first, geometry-driven approach, prioritizing deterministic behavior, reproducibility, and transparency over black-box inference.
CoreStereo is optimized for modern multi-core CPUs and supports scalable deployment across multiple stereo rigs.
- Host-side stereo processing (CPU-first)
- Full dense and ROI-based disparity estimation
- RAW-first, zero-copy pipeline design
- Deterministic, geometry-based reconstruction
- Multi-threaded and SIMD-optimized execution
- Scalable across multiple independent stereo nodes
- Optional GPU acceleration for advanced workloads
coming soon!