DeepCamera

Overview of DeepCamera

What it is

DeepCamera is an open-source AI camera skills platform for AI NVR and CCTV surveillance. It performs local VLM video analysis using models such as Qwen, DeepSeek, SmolVLM, LLaVA, and YOLO26. The platform includes an LLM-powered agentic security camera agent that watches, understands, remembers, and guards a home via Telegram, Discord, or Slack. It supports pluggable AI skills and can use OpenAI, Google, Anthropic, or local AI models. DeepCamera runs on Mac Mini and AI PC.

Key features

• Pluggable AI skills with a catalog covering detection, analysis, privacy, segmentation, annotation, training, automation, and integrations - Local VLM video analysis with models including Qwen, DeepSeek, SmolVLM, LLaVA, and YOLO26 - LLM-powered agentic security camera agent with persistent memory, agentic chat, AI video generation, voice (TTS), and conversational messaging via Discord, Telegram, or Slack - YOLO26 real-time object detection with multiple model sizes (nano, small, medium) and hardware acceleration (TensorRT, CoreML, OpenVINO, ONNX, Coral TPU)
• Hardware-aware installation that auto-detects GPU, NPU, or CPU and installs matching frameworks - Skill Store UI and AI/LLM-assisted skill installation - SharpAI Aegis desktop app for connecting cameras, deploying skills, and managing the AI pipeline

Use cases

• Home security surveillance with real-time object detection and VLM scene analysis - Person re-identification and face detection/recognition from camera feeds - Fall detection and CCTV/NVR monitoring - Privacy-preserving depth estimation to anonymize camera feeds while preserving activity - Dataset annotation and model training for custom detection models - Event-driven automation via MQTT, webhook, or Home Assistant triggers

Getting started

See the repository README for installation instructions.

License and pricing

The project is licensed under the MIT License and is open source.

Based on the project's public README and repository metadata.