Observation of migrating fish provides critical data required for recovery and management actions. Considerable resource is expended to count, speciate and sort migrating fish at purpose-built viewing facilities within dams and other man-made barriers. Manual operators observe and record the data in real time or post analyze video recordings. However, the data gathered, and decisions made are inherently prone to human error, operator fatigue and fish directional behavior. Turbidity can also exacerbate accuracy - the main reason that prior automation attempts have been largely unproductive. Recent development of machine vision technology used in manufacturing and fruit harvesting operations provides the potential for dramatically improving and simplifying fisheries data collection. In this session we describe an adaptation of the current state of the art to fisheries management. Using a simple false weir configuration, the fish are dewatered, singulated and descend a short, wetted slide. Controlled lighting and high-speed imagery from radially arranged cameras provide multiple photographs of consistent quality for real time processing. Using combinations of machine learning, image recognition and triangulation, the control system computers are able to simultaneously synthesize the needed data and provide signals for sorting actions in less than 2 seconds, with an extremely high degree of accuracy. Fish counts, and individual fork length and girth measurements can already be reliably captured. Currently under development are algorithms that include fin clip detection (for separation of wild and hatchery fish), and some speciation applications – primarily focused on exclusion of invasive species. The automated nature of the system facilitates 24-hour operation with real-time decisions and remote access to image data. Volitional fish passage is not interrupted, fish are not physically handled, spend minimal time de-watered and are efficiently classified allowing for selective passage.