Deploy AI models to edge devices to protect workers


This code pattern demonstrates how to deploy an AI workloads to an edge device for scoring.


In many factory environments, when employees enter a designated area, they must be wearing proper Personal Protective Equipment (PPE) such as a hard hat. This pattern demonstrates a solution which monitors the designated area and issues an alert only when an employee has been detected and is not wearing a hard hat. To reduce load on the network, the video stream object detection will be performed on edge devices that are managed by Open Horizon and IBM Edge Application Manager.

The trained models are containerized, stored in a model registry, and downloaded to the edge devices. The AI model runs on the edge device platform and performs worker safety object detection and edge prediction.

IBM Edge Application Manager is used to orchestrate the workloads to edge gateways or edge devices, and the IBM Edge Application Manager Agent starts and monitors those containerized workloads.



  1. AI Model developer uploads the training data to IBM Cloud Object Storage.
  2. Watson Machine Learning pulls the training data from IBM Cloud Object Storage and trains a model with TensorFlow. The trained model is saved back to IBM Cloud Object Storage.
  3. The trained models are containizered using Open Horizon and stored in a model registry.
  4. IBM Edge Application Manager orchestrates workloads to edge gateways or directly to edge devices (or both).
  5. IBM Edge Application Manager (Open Horizon) Agent invokes and monitors containizered workload.
  6. Models are downloaded from the model registry to the edge devices.
  7. AI model runs on the edge device to perform object detection and edge prediction.


Find the detailed steps for this pattern in the README file.



Related post