Article

Scientists at Oak Ridge National Laboratory (ORNL) have developed a method called Distributed Cross-Channel Hierarchical Aggregation (D-CHAG) that doubles processing speeds and reduces memory usage by 75% for analyzing plant imaging data. This advancement facilitates AI-guided research in developing high-yield crops and supports the Department of Energy's Genesis Mission.

D-CHAG enables efficient processing of hyperspectral imaging data from the Advanced Plant Phenotyping Laboratory (APPL) using the Frontier supercomputer. The method divides tasks among multiple graphics processing units (GPUs) and aggregates results in stages, allowing larger foundation models to be trained without losing data resolution.

Key results include substantial reductions in resource requirements and faster analysis capabilities, aiding in tasks like measuring plant photosynthetic activity. The project also contributes to broader agricultural innovations and enhances the ability to monitor crop health in real-time.