
Event Triggers and Notification
Event Triggers can be programmed based on inferencing jobs or workflows, and then OTA Programmed to a Smarter AI camera deployment or group. When an AI Container / AI model exceeds a classification threshold and/or a sensor exceeds a value, Smarter AI delivers Event Notifications via:
- Camera APIs
- Dashboard
- Mobile Apps / Libraries
- REST APIs
Benefits
AI Model Lifecycle Management
- Containerize and OTA program AI models
- Continuous learning / continuous delivery:
- Collect event notifications and recordings
- Retrain AI models – minimize false positives and negatives
- Containerize and program AI models OTA
- Train, containerize, and OTA program new AI models to support new use cases
Programmable Classification Thresholds
- OTA program classification thresholds
- Collect event notifications and recordings
- Optimize classification thresholds – minimize false positives and negatives
- Program a single AI Container with multiple classification thresholds
Real-time Event Notifications
- Receive event notifications via dashboard, mobile app and/or REST APIs
- Program the distribution of event notifications via Access Control Lists (ACL), Device Groups, and User Groups
- Configure event notifications with recordings, sensor values, and/or snapshots
Orchestration
- Provision and OTA program AI models
- Allocate Edge AI resources between multiple AI models
- Configure AI cameras and gateways in relation to AI models
- Monitor the health of AI models
Standard DeeaSupports standard deep learning frameworks and platforms
- Containerize and OTA program AI models trained with and for standard deep learning frameworks and platforms,
e.g. Caffe, CUDA, CVflow, MXNet, ONNX, OpenVINO, PyTorch, TensorFlow
Sensor Fusion
- Provision and OTA program AI models
- Field of view (FoV) from 0.15 to 360 degrees
- Image resolution from HD to 4K to 8K
- Object distance up to 50 miles
- Sensors including accelerometer, GPS, gyroscope, magnetometer, temperature
Inference Workflow Management
- AI models as programmable objects
- Chain inferencing jobs together with logical operators
- Map AI model outputs to AI model inputs
Inference Scheduling
- Inferencing on demand, continuous inferencing
- Run multiple AI models in parallel:
- using the same imager/sensor inputs
- using logical operators to infer decisions
Greenfield and Brownfield Deployments
- Containerize and OTA program AI models to greenfield (i.e. Smarter AI camera) deployments
- Containerize and OTA program AI models to brownfield (i.e. IP camera) deployments with Smarter AI Gateway
- Map local and distributed imager/sensor outputs to AI model inputs