Seminario Datalab
Recent advances in transfer learning and domain adaptation in computer vision
Ponente: Luis Herranz (UAM)Fecha: jueves 10 de abril de 2025 - 12:00Lugar: Aula Gris 1, ICMAT
Resumen:
The success of artificial intelligence (AI) is based on collecting and annotating huge amounts of data samples. However, in many cases this is too expensive or even impossible. A common solution is to transfer the knowledge stored in an already trained source model and adapt it to a target task using a small set of (often unlabeled) target samples. In this talk I will provide an overview of transfer learning and adaptation techniques, focusing on common settings such as domain adaptation and continual learning, and discriminative and generative scenarios in the context of computer vision.
