Detect persons, vehicles and potential security threats (such as knives, firearms, fire, chemical leakage, etc.) by utilizing motion detection to determine where to run object detection locally.
Multiple built-in detector cpu, edgetpu, onnx, openvino, rocm and tensorrt
High Inference Speed
Yolo-NAS-S | MobileDet
320 x 320
5 ms
Hardware Support
Nvidia GPU/Jetson
Intel Arc GPU | AMD GPU
Google Coral EdgeTPU
Customized Datasets
Train from scratch
No commercialization limits
Production-Ready
ReID
Use ReID (Re-Identification) to recognize the appearance characteristics of individuals and vehicles when facial features and license plates are not detectable.
Person
Head Hat, glasses
Upper color White, black, red, orange, yellow, pink, dark blue, blue, green, gray, purple
Upper Short sleeve, long sleeve, stride, logo, splice, plaid
Lower Stripe, pattern, long coat, shorts, trousers, skirts&dress, boots
Carrying Handbag, Shoulderbag, backpack, hold objects in front
Age Over60, 18-60, less18
Posture Front, side, back
Gender Female, male
Vehicle
Category Sedan, SUV, MPV, van, truck, pickup, bus, supercar
Model Xiaomi·su7, Model-3|Model-Y, Passat·2019-2021, AITO M9, MEGA and more than 400 models
Semantic Search
By generating embeddings—numerical vector representations—for both the images and text descriptions of your tracked objects, the system compares these embeddings to evaluate their similarities and provide relevant search results.
Similarity Search
The vision model Jina AI CLIP is able to embed both images and text into the same vector space, which allows image -> image and text -> image similarity searches.
Generative AI
Generative AI—powered by self-hosted large language models via Ollama, Google Gemini, and OpenAI—can automatically generate descriptive text based on the thumbnails of your tracked objects. This enhances Semantic Search by providing richer context about your tracked objects. Beyond simply recognizing 'what' is in a scene, it aims to infer 'why' it might be there or 'what' it could do next.
AIoT
Enable IoT connectivity through the MQTT protocol.
Integration with Home Assistant via plug-ins allows for linkage control of relays, lighting, appliances, UPS systems, and more.
The system provides an MQTT topic, which is updated with a JSON payload containing the event_id and description whenever your AI provider generates a description for a tracked object. This description can be directly utilized in notifications, such as sending alerts to your phone or triggering audio announcements.
| Computer Vision Solutions
Monitor Any Security-Critical Scenario
The models and docker images are free to use (already in Github repository and Docker Hub)
| Computer Vision Solutions 
