Perceive
Camera, microphone and text enter through separate adapters and CUDA preprocessing.
Sensory is an experimental CUDA platform for stateful neural systems. It accepts images, audio, text or raw vectors and lets them act on a persistent recurrent network described in Cognia.
Inputs and outputs are adapters around the same live neural graph.
Camera, microphone and text enter through separate adapters and CUDA preprocessing.
Activation, adaptation, delayed edges, attention and workspace persist between steps.
Declared output regions become signals or text available in the application and over HTTP.
Available means implemented in the repository, not production-guaranteed.
Multimodal preprocessing, Cognia compilation, stateful and stateless inference, supervised readouts and observability.
Global workspace, plasticity, chemistry and structural growth are implemented for research.
Multiple simultaneous networks and runtime event pub/sub are not fully bridged yet.
Explore adaptive control, sequence classification, working memory or attention with an explicit baseline and measurable outcome.