This article describes how Enel, a multinational energy company, uses Amazon SageMaker and computer vision to automate power grid asset management and anomaly detection across its 2.3 million kilometer distribution network.

• Enel processes 40+ million images annually using deep learning models for infrastructure inspection
• Built ML factory platform on SageMaker for training models across multiple business units
• Uses three data sources: LiDAR point clouds, high-resolution aerial images, and satellite imagery
• Employs EfficientNet, EfficientDet, and SimCLR architectures for classification and reidentification tasks
• LiDAR point clouds enable precise 3D measurements with centimeter-level accuracy
• Satellite imagery monitors vegetation density to optimize tree pruning operations
• SageMaker Training enables parallel GPU-accelerated model training without infrastructure management
• Hierarchical pipeline identifies regions of interest, classifies assets, detects anomalies, and clusters duplicate pole images

Enel leverages SageMaker's scalability to automate expensive manual inspection processes, improving efficiency and accuracy for managing critical power infrastructure serving 60+ million customers.