Expand description
Framework-agnostic vision interfaces for CortenForge detection pipelines.
This crate defines the core abstractions for detection, capture, and recording:
Detector: Trait for running inference on frames.Recorder: Trait for persisting frame records and labels.Frame,FrameRecord,Label: Core data types for vision pipelines.- Capture resources and overlay utilities.
§Design Philosophy
vision_core is intentionally framework-agnostic. It has no Bevy dependencies and can be
used in CLI tools, web services, or any Rust application. The vision_runtime crate
provides Bevy-specific integration on top of these interfaces.
§Stability
The core interfaces (Detector, Recorder, FrameSource) and data types (Frame, Label,
DetectionResult) are stable and follow semantic versioning. Breaking changes to these
types will result in a major version bump.