Currently, my research interests include several themes in the area of 3D Computer Vision, some of them developed in collaboration with other Institutions (see also my active research collaborations below):

  • FACE RECOGNITION FROM 2D AND 3D DATA - I am interested in "in the wild" face recogniton solutions, in particular, "open set" and "open world" face recognition using CNNs and GANs
  • 3D MORPHABLE MODELS AND THEIR USE IN 2D/3D FACE RECOGNITION - We develop a 3DMM solution based on dictionary learning (DL-3DMM) that solve most of the limitations of hte standard method (i.e., local deformations, expression generation, etc.)
  • FACIAL EXPRESSION AND EMOTION RECOGNITION - We published several works on facial expression recognition from 3D static and 3D dynamic data. I am still actively investigating this area of research looking to solutions for facial expression recognition in videos that combine the power of deep learning with the modeling of trajectories on well defined geometric manifolds
  • FACE RECONSTRUCTION FROM LOW-RESOLUTION DEPTH DATA - We provided several contributions on this topic following the idea of reconstructing a higher resolution face model from a sequence of low-resolution frames acquired with a Kinect camera. Currently, we are working on solutions that allow long-term reconstruction of the face in uncooperative contexts, with applications in face recognition and person re-identification
  • ACTION RECOGNITION AND ACTION PREDICTION FROM RGB-D DATA - We proposed several methods for detecting and recognizing human actions from skeletal data. Some of them had large impact on the litarature in the field. Some innovative ideas were related to the modeling of the movement as trajectories on some non-linear manifolds. 
  • SURFACE DESCRIPTORS FOR RELIEF PATTERNS REPRESENTATIONS ON 3D MESHES - We were the first to provide an effective extension of LBP to the mesh manifold domain by defining meshLBP. meshLBP has found applications in 3D face recogniton, 3D relief pattern classification and retrieval. Currently, we are working on the idea of extending the proposed method to most of the LBP variants, while we are also generalizing it to different filtering operations on the mesh
  • EXTENSION OF CNN ARCHITECTURES TO NON-EUCLIDEAN DOMAINS - I am interested in models for redefining CNN operations on mesh manifold domains
International Collaborations
Since 2018: Prof. Anup Basu, Faculty of Science, University of Alberta, Edmonton, Alberta, Canada

Since 2018: Prof. Aparecido Nilceu Marana, Departamento de Computacion Faculdade de Ciencias (UNESP), Campus de Bauru, Brazil. One year scholarship (august 2018/july 2019) of a Ph.D. student at MICC-UNIFI 

Since 2015: Prof. Bjorn Ottersten and Dr. Dijamila Aouada, Centre for Security, Reliability and Trust (SnT), University of Luxembourg, Luxembourg. Supervision committee of several Ph.D. students at SnT
Since 2012: Prof. Naoufel Werghi, Khalifa University of Science, Technology and Research (KUSTAR), Abu Dhabi, UAE. Common research on the theme of surface description for relief patterns using mesh-LBP. Supervision of a Ph.D. student at KUSTAR
Since 2009: Prof. Mohamed Daoudi, IMT Lille Douai, Ecole Mines Télécome, IMT-Université de Lille, Lille, France. Several shared direction and co-supervision of Ph.D. students on the theme of 3D facial expression recognition and human action recognition from RGB-D data
Since 2006: Dr. Francisco Josè Silva Mata, Advanced Technologies Application Center (CENATAV), La Habana, Cuba. Common research on the topic of 2D/3D and 3D face recognition.

From 2005 to 2011: Prof. Antonio Mosquera, University de Santiago De Compostela, Spain. Co-supervison of Ph.D. students