Fishery-Independent Surveys and Machine Learning for Stock Assessment скачать в хорошем качестве

Fishery-Independent Surveys and Machine Learning for Stock Assessment

Presentation by Benjamin L. Richards, Stock Assessment Program, Fisheries Research and Monitoring Division, NOAA Fisheries Pacific Islands Fisheries Science Center at Video and Image Analytics (VIA) Task Force monthly meeting on August 15, 2019. Abstract: "Optical, camera-based surveys confer myriad advantages to the marine scientist, including non-extractive sampling with consistent bias and creation of auditable datasets. Unfortunately, camera-based surveys currently generate data volumes that quickly exceed capabilities of human annotators. NOAA Fisheries has been engaged in a five-year Strategic Initiative to support development of machine learning and computer vision algorithms to automate processing of optical datasets generated through marine resource surveys. In 2018, NOAA Fisheries released two software packages that support automated annotation, optimizing human annotation efforts and increasing efficiency of data processing."