Learning to Adapt: Domain Adaptation for Visual Recognition


Deep networks have significantly improved the state of the arts for several tasks in computer vision. Unfortunately, the impressive performance gains have come at the price of the use of massive amounts of labeled data. As the cost of collecting and annotating data is often prohibitive, given a novel target task where few or no training samples are available, it would be desirable to build effective learners that can leverage information from labeled data of a different but related source domain. However, a major obstacle in adapting predictive models to the target task is the shift in data distributions across different domains. This problem, typically referred as domain shift, has motivated research into Domain Adaptation (DA).

Traditional DA algorithms assume the presence of a single source and a single target domain. However, in real-world applications different situations may arise. For instance, in some cases multiple datasets from diverse source domains may be available, while in other settings target samples may not be given at the training stage or could arise from temporal data streams. Alternatively, in some applications knowledge about different domains may be encoded only in form of side-information (e.g. metadata, text) and should be effectively exploited to guide the adaptation process. In the last few years, we developed several deep learning-based approaches for DA for visual recognition and robot perception tasks.

Relevant Publications:
  • M. Mancini, H. Karaoguz, E. Ricci, P. Jensfelt, B. Caputo, Knowledge Is Never Enough: Towards Web Aided Deep Open World Recognition, ICRA 2019.
  • M. Mancini, S.R. Bulò, B. Caputo, E. Ricci, AdaGraph: Unifying Predictive and Continuous Domain Adaptation through Graphs, CVPR 2019. (Spotlight Oral, 8% acceptance rate)
  • S Roy, A Siarohin, E Sangineto, SR Bulo, N Sebe, E Ricci, Unsupervised Domain Adaptation using Feature-Whitening and Consensus Loss, CVPR 2019.
  • M. Mancini, L. Porzi, S. Rota Bulò, B. Caputo, E. Ricci: Boosting Domain Adaptation by Discovering Latent Domains. CVPR 2018. (Spotlight Oral, 8% acceptance rate)