Keynote Talk 2: Dr. Oisin Mac Aodha

Knowledge Discovery in Fine-grained Image Collections

Visual data collected and distributed on citizen science provides a wealth of information related to the natural world. Over the past few years, we have witnessed impressive advances in our ability to automatically process such data, e.g., we can now perform tasks such as large-scale species classification to a relatively high level of accuracy. However, identifying existing known species in such data is only the tip of the iceberg. These datasets contain a vast amount of other information that is currently not being extracted. In this talk, I will present some recent work from my group where I will outline new methods we are developing for automatically discovering concepts in large image collections and empowering researchers to efficiently search images for ecological signals of interest.