Piacentino, M., & Zhang, D. (2017, August). Automated Image Analysis and Classification Tool Based on Computer Vision Deep Learning Technologies. In 147th Annual Meeting of the American Fisheries Society. AFS.
Abstract
Today, underwater fishery surveys collect very large sets of digital video but extensive manual efforts are required to perform underwater image analysis of these videos. We present a rapid underwater video and automated image analysis tool using computer vision deep learning technologies to provide both (1) rapid semi-automated object annotation and (2) object detection and classification (e.g. types of fish, plant life, coral, rock, etc.) to support rapid analysis of underwater conditions.
During the development of our automated image analysis tool, developed for NOAA, we learned annotated data from video is often very limited due to effort required, therefore a tool that first provides annotation methods to train computer vision algorithms is greatly needed. Through our semi-automated, analyst guided annotation tool, rapid annotation methods are now available to enable annotation to setup follow-on automated analysis tools.
Our tool has been integrated into the NOAA VIAME (Video and Image Analytics in a Marine Environment) open-source video processing framework for research analysts. The tool, through its analyst guided annotation capabilities, can support analysis for all trained fish types, plant life, terrain types or identification of anomalous items allowing researchers to process new and existing videos, yielding desired analysis and classification results.