Deploying computer vision systems - A case study on birdsong identification

Speaker: Grant Van Horn , UMass Amherst

Date: Thursday, December 07, 2023

Time: 1:00 PM to 2:00 PM Note: all times are in the Eastern Time Zone

Public: Yes

Location: 2-190

Event Type:

Room Description:

Host: Sara Beery, Phillip Isola, Jeremy Bernstein, CSAIL MIT

Contact: Thien Le, thienle@csail.mit.edu

Relevant URL:

Speaker URL: None

Speaker Photo:
None

Reminders to: seminars@csail.mit.edu, mitml@mit.edu, lids-seminars@mit.edu, vc-all@csail.mit.edu, ei-all@lists.csail.mit.edu

Reminder Subject: TALK: Deploying computer vision systems - A case study on birdsong identification

Abstract:
The Merlin Bird ID app integrates Merlin Sound ID, a sophisticated system designed for real-time classification of bird vocalizations on mobile devices. With a simple tap, your phone's microphone becomes a gateway to the symphony of bird song, as Sound ID reveals the hidden identities of avian artists in your surroundings. In this discussion we will delve into the critical components of this application, including the data underpinning it, the annotation processes, and an overview of our training protocols. We'll also explore our benchmarking procedures and ongoing efforts to enhance performance. This talk aims to shed light on the challenges and solutions associated with deploying real-time auditory classification systems in ecological applications.
Bio:
Grant Van Horn is an Assistant Professor in the Manning College of Information and Computer Sciences at the University of Massachusetts, Amherst, and a visiting researcher at the Cornell Lab of Ornithology. His research lies at the intersection of computer vision and machine learning, with an emphasis on crafting real-world machine learning systems that integrate human expertise, state-of-the-art machine learning methodologies, and large-scale datasets. Merlin Sound ID is his latest contribution in this space, following the success of Seek, the iNaturalist computer vision system, and Merlin Photo ID. He completed his PhD at Caltech in 2019, advised by Pietro Perona. His thesis work focused on efficient dataset collection through human-in-the-loop systems, and fine-grained visual categorization. He completed his BS and MS at UCSD where he was advised by Serge Belongie, and his work falls under the broad research agenda of Visipedia.

Research Areas:
AI & Machine Learning

Impact Areas:

This event is not part of a series.

Created by Thien Le Email at Wednesday, December 06, 2023 at 12:18 PM.