F
Fabio Maria Carlucci
Researcher at Sapienza University of Rome
Publications - 31
Citations - 1647
Fabio Maria Carlucci is an academic researcher from Sapienza University of Rome. The author has contributed to research in topics: Deep learning & Domain (software engineering). The author has an hindex of 15, co-authored 31 publications receiving 1087 citations. Previous affiliations of Fabio Maria Carlucci include Istituto Italiano di Tecnologia & Huawei.
Papers
More filters
Proceedings ArticleDOI
Domain Generalization by Solving Jigsaw Puzzles
TL;DR: This model learns the semantic labels in a supervised fashion, and broadens its understanding of the data by learning from self-supervised signals how to solve a jigsaw puzzle on the same images, which helps the network to learn the concepts of spatial correlation while acting as a regularizer for the classification task.
Proceedings ArticleDOI
From Source to Target and Back: Symmetric Bi-Directional Adaptive GAN
TL;DR: In this paper, a symmetric mapping among domains is proposed to preserve the class identity of an image passing through both domain mappings, and a new class consistency loss is defined to align the generators in the two directions.
Posted Content
NAS evaluation is frustratingly hard
TL;DR: This work proposes using a method’s relative improvement over the randomly sampled average architecture, which effectively removes advantages arising from expertly engineered search spaces or training protocols to overcome the hurdle of comparing methods with different search spaces.
Proceedings Article
NAS evaluation is frustratingly hard
TL;DR: In this article, the authors compare 8 NAS methods on 5 datasets and find that the cell-based search space has a very narrow accuracy range, such that the seed has a considerable impact on architecture rankings; the hand-designed macrostructure (cells) is more important than the searched micro-structure.
Posted Content
AutoDIAL: Automatic DomaIn Alignment Layers
TL;DR: Opposite to previous works which define a priori in which layers adaptation should be performed, this method is able to automatically learn the degree of feature alignment required at different levels of the deep network.