Equitus Video Sentinel (EVS) development pipeline

Pilot Proposal:  Customizing and managing VMS Integration for IBM Power 11  

Equitus demonstrates the value of AI-enhanced video in regulated industries by using a structured development pipeline that integrates video analytics with enterprise systems like IBM Power 11 and Maximo. The pipeline uses specialized tools for each stage, from data annotation to model deployment.

Data Annotation and Preparation

Equitus uses three distinct annotation tools to prepare high-quality datasets for training its AI models.

• CVAT (Computer Vision Annotation Tool): This is the primary tool for video-centric projects. Its collaborative features and support for detailed video annotation formats (like bounding boxes and polygons) are crucial for labeling objects and events in surveillance or operational footage for auditing and compliance purposes.

• Label Studio: This tool provides a more flexible solution for projects that may involve various data types beyond video, such as text from maintenance logs or audio recordings, allowing for the creation of multi-modal datasets.

• VGG Image Annotator (VIA): A lightweight, browser-based tool used for fast, one-off image annotation tasks, likely for quick prototyping or smaller, less complex datasets.

Model Deployment with ONNX

ONNX (Open Neural Network Exchange) is the critical link between model training and deployment. The data annotation tools themselves do not output ONNX models. Instead, they produce annotated data, which is then used to train a model in a framework like PyTorch or TensorFlow. Once trained, the model is converted to the standardized ONNX format.

• Interoperability: ONNX's main benefit is its ability to run models on a wide range of hardware and software, including the specialized IBM Power 11 architecture. This ensures that a model trained on one system can be efficiently deployed on others without losing accuracy.

• Auditability and Compliance: Using a standardized format like ONNX provides a consistent and reproducible deployment process, which is essential for regulated industries that require clear audit trails and verifiable model performance.

---

Integration with IBM Power 11 and Maximo

The final value proposition lies in the integration of the AI-enhanced video insights with IBM's enterprise platforms.

• IBM Power 11: The ONNX-formatted AI models are deployed on IBM Power 11 servers. These servers are designed for high-performance computing and are well-suited to handle the computationally intensive task of real-time video analysis for security and compliance. The reliability and security of this hardware are key for mission-critical applications.

• IBM Maximo: The insights generated by the AI models (e.g., detecting a safety violation, identifying a broken machine part, or noting a compliance issue) are automatically fed into the Maximo Enterprise Asset Management (EAM) system. This integration allows the AI to trigger specific actions within Maximo's workflow, such as:

  • Automated Auditing: Automatically flag video events that require human review, creating a digital audit trail.

  • Proactive Maintenance: Identify potential equipment failures from video analysis and automatically create a work order in Maximo.

  • Security Alerts: Trigger alerts in Maximo's security module for unauthorized access or other suspicious activity.

By automating these processes and connecting them to existing enterprise systems, Equitus helps regulated industries improve operational efficiency, enhance security, and ensure compliance with a verifiable, AI-driven workflow.