BEGIN:VCALENDAR VERSION:2.0 PRODID:-//128.220.36.25//NONSGML kigkonsult.se iCalcreator 2.26.9// CALSCALE:GREGORIAN METHOD:PUBLISH X-WR-CALNAME:Laboratory for Computational Sensing + Robotics X-WR-CALDESC: X-FROM-URL:https://lcsr.jhu.edu X-WR-TIMEZONE:America/New_York BEGIN:VTIMEZONE TZID:America/New_York X-LIC-LOCATION:America/New_York BEGIN:STANDARD DTSTART:20231105T020000 TZOFFSETFROM:-0400 TZOFFSETTO:-0500 RDATE:20241103T020000 TZNAME:EST END:STANDARD BEGIN:DAYLIGHT DTSTART:20240310T020000 TZOFFSETFROM:-0500 TZOFFSETTO:-0400 RDATE:20250309T020000 TZNAME:EDT END:DAYLIGHT END:VTIMEZONE BEGIN:VEVENT UID:ai1ec-13115@lcsr.jhu.edu DTSTAMP:20240328T132907Z CATEGORIES: CONTACT:Ashley Moriarty\; amoriar2@jhu.edu DESCRIPTION:
\n
Abstract:
\nI will present a bio- inspired fish simulation platform\, which we call “Foids”\, to generate re alistic synthetic datasets for an use in computer vision algorithm trainin g. This is a first-of-its-kind synthetic dataset platform for fish\, which generates all the 3D scenes just with a simulation. One of the major chal lenges in deep learning based computer vision is the preparation of the an notated dataset. It is already hard to collect a good quality video datase t with enough variations\; moreover\, it is a painful process to annotate a sufficiently large video dataset frame by frame. This is especially true when it comes to a fish dataset because it is difficult to set up a camer a underwater and the number of fish (target objects) in the scene can rang e up to 30\,000 in a fish cage on a fish farm. All of these fish need to b e annotated with labels such as a bounding box or silhouette\, which can t ake hours to complete manually\, even for only a few minutes of video. We solve this challenge by introducing a realistic synthetic dataset generati on platform that incorporates details of biology and ecology studied in th e aquaculture field. Because it is a simulated scene\, it is easy to gener ate the scene data with annotation labels from the 3D mesh geometry data a nd transformation matrix. To this end\, we develop an automated fish count ing system utilizing the part of synthetic dataset that shows comparable c ounting accuracy to human eyes\, which reduces the time compared to the ma nual process\, and reduces physical injuries sustained by the fish.
\n< p> \nBio: Masaki Nakada obtained a master degree in physics at Wase da University in Japan. Then\, he finished PhD in computer science at UCLA and worked as a postdoc for another year\, where he published a series of scientific papers. (https://w ww.masakinakada.com/) He devoted more than 10 years in the research of artificial life\, specifically in the area of biomechanical human simulat ion with musculoskeletal models\, neuromuscular controllers\, and biomimet ic vision. Previously\, he worked for Intel as a software engineer. He rec eived MIT Technology Review Innovator Award Under 35\, Forbes Next 1000\, Institute for Digital Research and Education Postdoctoral Scholar Award\, Siggraph Thesis Fast Forward Honorable mention\, TEEC Cup North American Entrepreneurship Competition in Silicon Valley\, Japan Student Services Or ganization Fellowship\, Rotary Ambassadorial Fellowship\, Itoh Foundation Fellowship\, Entrepreneurship Foundation Fellowship\, Aoi Foundation Fello wship and winner of several Startup business competition & hackathons. He founded NeuralX\, Inc (https://www.neu ralx.ai/) in 2019 based on the IP he has developed over the decade of research. The company provides an interactive online fitness service Prese nce.fit (https://www.presence.fit/ )\, where it combines the power of human instructor and motion analytics A I\, which enables them to provide highly interactive online fitness experi ence.
DTSTART;TZID=America/New_York:20221026T120000 DTEND;TZID=America/New_York:20221026T130000 LOCATION:Hackerman B17 SEQUENCE:0 SUMMARY:LCSR Seminar: Masaki Nakada “Foids: Bio-Inspired Fish Simulation fo r Generating Synthetic Datasets” URL:https://lcsr.jhu.edu/events/masaki-nakada/ X-COST-TYPE:free END:VEVENT END:VCALENDAR