Prof. Nicu. Sebe (Univ. of Trento, Italy)
Title: Deep Learning for Analysis and Generation of Facial Attributes.
Abstract: Face analysis is a fundamental and challenging problem in computer vision. For face analysis, one of the fundamental techniques is face alignment which is usually employed as the preprocessing step. For instance, when we do face verification, the faces need to be aligned first. On top of that, face pose and facial expression normalization is also a very important preprocessing step. In addition to the disturbance of the poses and expressions, another very important factor that might degrade the performance of face verification is age. Thus, modeling the aging process of human faces is important for cross-age face verification and recognition. Similar to face verification, a face aging task also needs to normalize the face poses and expressions, and many works have been proposed to normalize for facial expressions. However, only a few works have studied how to generate faces having various expressions, such as spontaneous or posed smiles. In this presentation, we will present our recent results on face alignment, face aging, and smile video generation employing recurrent neural networks, such as the Long Short Term Memory (LSTM) and the Gated Recurrent Unit (GRU).
Short Bio: Nicu Sebe is a professor in the University Of Trento, Italy, where he is the director of the Department of Information Engineering and Computer Science. He is leading the research in the areas of multimedia information retrieval and human-computer interaction in computer vision applications. He was involved in the organization of the major conferences and workshops addressing the computer vision and human-centered aspects of multimedia information retrieval, among which as a General Co-Chair of the IEEE Automatic Face and Gesture Recognition Conference, FG 2008, ACM International Conference on Image and Video Retrieval (CIVR) 2007 and 2010. He was a general chair of ACM Multimedia 2013 and ACM ICMR 2017 and a program chair of ACM Multimedia 2011 and 2007, ECCV 2016 and ICCV 2017. He is the program chair of ICPR 2020. Currently he is the ACM SIGMM vice chair. He is a fellow of IAPR and a Senior member of ACM and IEEE.
Prof. Ricardo Chavarriaga (EPFL STI IBI-STI CNBI Switzerland )
Title: Symbiotic interaction through brain-machine interfacing, machine learning and VR.
Abstract: Brain-machine interfaces (BMI) decode neural activity into commands used in clinical applications (e.g. neuroprosthetics for restoring or substitute lost motor capabilities), as well as consumer applications (e.g. gaming, wellness or neuroergonomics). Since this interfaces are expected to operate in real-life situations, research in BMI is required to develop methods able to process neural activity in a single-trial basis in less- controlled scenarios than typical protocols used in laboratory conditions. In consequence, these methods provide a powerful tool to evaluate neural processes mediating motor and cognitive capabilities in realistic situations. Advances in both machine learning and virtual reality offer a great opportunity to further advance this field, leverage our knowledge of the nervous system and develop more robust human-machine interfaces. Nonetheless, this opportunity comes with multiple challenges both technical and methodological. For instance, despite impressive achievements of current trends such as deep learning in other fields, their success has not been translated to BMI applications. Here I will discuss some of the these challenges and strategies to overcome them.
Short Bio: Ricardo Chavarriaga is a senior researcher at the Center of Neuroprosthetics of the Ecole Polytechnique Fédérale de Lausanne (EPFL), Switzerland. He holds a PhD in computational neuroscience and has more than 100 scientific publications on brain-machine interfacing and neurotechnologies. His work focuses on understanding and decoding the neural mechanisms that govern cognitive processes such as error awareness, attention and decision making. His vision targets a future where this information can be used to endow artificial devices with the capability to adapt to their user’s goals, preferences and capabilities so as to establish a truly symbiotic relation between human and machines.
Website: http://people.epfl.ch/ricardo.chavarriaga. Twitter: @Chavarriaga_BCI
Prof. Jordi González (CVC-UAB, Spain)
Title: Going beyond Deep Learning for Understanding Human Behaviours in Image Sequences.
Abstract: Although the recently re-discovered field of deep learning has revolutionized areas like computer vision, such approaches have still several limitations that should be kept in mind in order to design more advanced artificial general intelligent systems for human behavior understanding and human-computer interaction, among other complex tasks. In particular for those two domains, there indeed is a potentially infinite range of input motions with a potentially infinite range interpretations. In this talk, we will cover how the combination of rule-based symbolic reasoning, prior knowledge and common-sense integration, and high-level abstract concepts manipulation can provide more insights to properly interpret the behaviors of humans in front of a camera. If concepts like space, time and object can be represented, the use of hierarchical symbolic systems used for inference can strongly benefit from the so powerful classification performance achieved by neural networks. As a result, by considering deep learning as just part of a more complex and more challenging problem-solving system, a better understanding of human motion can be achieved, including abstraction, structure, intention, open-world and discovery capabilities.
Short Bio: Jordi Gonzàlez received the Ph.D. degree in computer engineering from Universitat Autònoma de Barcelona (UAB) in 2004. He is at present associate professor in computer science at the Computer Science Department, UAB. He is also senior researcher at the Computer Vision Center, where he has co-founded three spin-offs (Cloud Size Services, Visual Tagging, and Care Respite) and the Image Sequence Evaluation (ISE Lab) research group. His research interests include machine learning techniques for the computational interpretation of human behaviour in social images, or Visual Hermeneutics.
Prof. José Luis Lisani (DMI-UIB, Spain)
Title: Color Preprocessing and its Effect on Learning.
Abstract: In this talk we will discuss the role of color preprocessing in the performance of CNNs used for classification tasks. The goal of preprocessing will be to obtain a normalized representation of color images that will make the classification task robust to variations in the lighting conditions. This preprocessing will be applied in the training stage, where all the samples will be color-normalized, but also in the classification of any input sample. We shall investigate how this normalization affects the performance of the designed CNN when applied to samples coming from datasets different from the one used for training.
Short Bio: Jose-Luis Lisani received the PhD degree in Computer Science and Applied Mathematics by
the Universities of Illes Balears (Spain) and Paris-Dauphine (France) in 2001. His research interests include the analysis and processing of color images and video sequences. For more than ten years he has collaborated with the US-based forensic software company Cognitech Inc. and he is co-inventor of 8 US patents. He has co-authored the book “A Theory of shape identification” (Springer LNM, 2008).
He is currently Assistant Professor at the University of Illes Balears (Spain).