This paper presents two use cases where deep learning is able to help scientists by removing the burden of manual review of large volumes of physical data. Such examples highlight why deep learning could become a transverse tool across many scientific fields.