TypeCV documentation

TypeCV covers computer vision broadly. SfM is a current implementation area within geometry and 3D vision. Planned capabilities and implemented APIs are identified separately throughout these documents.

Project-wide

• Project introduction and development
• Architecture and TypeShade execution model
• TypeShade capabilities: source review and executable GPU probes
• GPU geometric vision operators, resource contracts and optimization
• Resident GPU pair verification, E/H estimation and relative pose
• Computer vision roadmap and evidence gates

Geometry and 3D vision: current SfM implementation

• SfM quickstart and current scope
• Calibrated two-view API and numerical behavior
• Incremental sparse reconstruction and bundle adjustment
• TypeShade GPU execution and resident scoring
• SfM-specific roadmap

Experiments

• Exploratory two-view pyramid video experiment
• Controlled-input pyramid video / COLMAP comparison
• Recorded software WebGPU run
• GPU P3P branch comparison on pyramid observations
• GPU PnP/triangulation/BA in the incremental pyramid reconstruction
• GPU numerical initialization through BA, with similarity-aligned camera comparison
• Resident pair verification: identical models, measured maps/bytes and buffer tradeoffs

Additional domain guides will accompany their implementations; the roadmap does not imply that image processing, optical flow, SLAM or learned-vision APIs are already available.