The intelligent vision layer for the Linux of AI. Profile-driven analysis, real-time processing, zero data egress.
Perceptor turns your existing camera feeds into intelligent situational awareness. It detects human presence—not just motion—using duration, context, and profile-based behavior analysis. No cloud, no APIs, no subscriptions. Everything runs 100% on your hardware, delivering selective, contextual alerts only when presence matters.
Designed for construction sites, retail environments, and facilities where cameras already exist—but nobody is watching at 3 a.m.
Eight intelligence modules. One unified presence awareness engine. Running entirely on your hardware with profile-driven logic.
Distinguishes human presence from environmental motion with profile-aware behavior analysis. Construction, retail, or facility modes available.
Human body & hand pose tracking to understand posture, intent, and anomalous behavior patterns for contextual alerting.
On-device biometric identification for authorized personnel tracking and unknown presence detection.
Extracts and reads text from visual scenes—vehicle plates, badges, signage—for enriched situational context.
Speech-to-text and voice command integration. Links with Blackbox Audio for complete acoustic awareness.
Multimodal reasoning with Qwen3-30B-VL and similar models. Interprets visual scenes with language understanding locally.
Large Language Models evaluate presence context and generate selective, profile-aware alerts—no API calls, pure local compute.
High-performance async processing engine for real-time, non-blocking sensor fusion and inference across all perception modules.
Built with the same technologies powering AetherPro's industrial AI infrastructure.
Perception kernel and four modules stable. Profile configuration system in progress. Target: Q1 2026 public release.
For operations teams and security managers deploying sovereign AI perception. Limited pilot slots.