CVC is more than happy to announce that three of its papers have been selected for the ICCV2021 Conference, the premier international computer vision event.
Two of them were from Joost van de Weijer’s group and the other one from Sergio Escalera’s. Hereafter you can find them with a summary.
– TransferI2I: Transfer Learning for Image-to-Image Translation from Small Datasets
Yaxing Wang, Hector Laria Mantecon, Joost van de Weijer, Laura Lopez-Fuentes, Bogdan Raducanu
Learning image-to-image (I2I) translation requires large datasets. In this paper, the authors propose transfer learning theory for I2I and show that they can now train high-quality systems on very small datasets.
– Generalized Source-free Domain Adaptation
Shiqi Yang, Yaxing Wang, Joost van de Weijer, Luis Herranz, Shangling Jui
The authors propose a new method that can adapt to new target domains while keeping performance on the source domain without the need for any labels on the target domain.
– DeePSD: Automatic Deep Skinning And Pose Space Deformation For 3D Garment Animation
Hugo Bertiche, Meysam Madadi, Emilio Tylson, Sergio Escalera
The authors present a novel solution to the garment animation problem through deep learning. Their contribution allows animating any template outfit with arbitrary topology and geometric complexity. When giving a piece (or set) of clothes, our approach links it to a “human body skeleton” thus, both clothing and person move in unison. Additionally, small deformations to the outfit for a realistic and consistent result are added.