Lecturers
Henry Arguello Fuentes
Ph.D. Electrical and Computer Engineering, University of Delaware
Associate Professor, Universidad Industrial de Santander.
Abstract
Compressive Spectral Imaging (CSI) technique senses the spatio-spectral information of a scene by measuring 2D coded projections on a focal plane array. Several coded spectral imager systems have been recently developed realizing CSI, such as the Coded Aperture Snapshot Spectral Imager (CASSI) and the Prism-Mask Video Imaging Spectrometry (PMVIS). Moreover, the reconstruction image quality of the optical CSI systems directly depends on the design of a 2D set of binary coded apertures, which block-unblock the light from the underlying scene. Therefore, these conferences present several mathematical approaches to obtain optimal coded aperture designs in CSI systems, in order to improve the quality of the reconstructed spectral images. Further, extensive simulations show the improvement of the different proposed coded aperture designs in terms of the Peak signal-to-noise ratio (PSNR). The conference starts with an introduction to the basic principles of Compressive Sensing, Spectral Imaging, and Compressive Spectral Imaging where the optical compressive architectures are presented. After, different techniques as Principal Component Analysis, Incoherence and Spatial Filtering are presented to design the coded apertures. Also, applications of spectral imaging classification and super-resolution are presented.
​
Bio
Henry Arguello is an IEEE Senior member. He earned the Ph.D. degree in Electrical and Computer Engineering at the University of Delaware in 2013. Since 2006, he has served as professor in the Department of Systems Engineering at Universidad Industrial de Santander, Colombia. Additionally, he leads the High Dimensional Signal Processing (HDSP) research group. His research interests include high-dimensional signal coding and processing, optical imaging, compressed sensing, hyperspectral imaging, optical coding, computarional imaging, numerical optimization, and stochastic algorithms. Results of his research have been published in more than 50 papers in high impact scientific journals including IEEE Signal Processing Magazine, IEEE Trans. on Image Processing, IEEE Trans. on Geoscience and Remote Sensing, Applied Optics Express, Journal of the Optical Society of America A, and Optical Engineering.
​
Lectures
1. Introduction to compressive sensing.
2. Spectral Imaging (Introduction, Applications).
3. Compressive Spectral Imaging.
4. Random pattern design in Compressive Spectral Imaging.
Abstract
This talk will be composed of two parts. Initially, the basics for obtaining information about the geometry of 3D scenes (3D reconstruction) by using 2D image (monocular system) are presented. Particularly, the technique known as Structure from Motion (SFM) will be studied, for which the following aspects will be described: Pinole projection model; Camera calibration and the projection matrix; Corresponding points in two views; Fundamental matrix; Essential matrix; 3D points by triangulation. Also a tool for 3D reconstruction will be described. Later, different techniques related to the process of 3D point clouds (resulting from the 3D reconstruction), will be illustrated by developing two applications in outdoor environments: The detection of electric towers in rural scenes and the classification of vegetative structures on coffee branches. Particularly, for these clouds of points, will be addressed aspects such as: Filtering; Key points; Descriptors; Segmentation and Classification. For this second part, the processing of the 3D point cloud will be done using the Point Cloud Library (PCL).
​
Bio
Prof. Prieto received a degree in electronic engineering in 1991 from Francisco José de Caldas Distrital University, a degree in physics in 1992 from the National University of Colombia, and his M.Sc. in electrical engineering in 1995 from Andes University, all in Bogotá, Colombia. He received his Ph.D. in 2000 in industrial automation from the Institut National des Sciences Appliquées at Lyon, France and from the École de Technologie Supérieure, Université du Québec, Montréal, Canada. His Ph.D. dissertation was developed in cooperation between the two universities. He joined the National University of Colombia in February 2000 as a professor. Since his PhD studies and as a professor, his research activities have been focused on computer vision (2D and 3D), image processing, automated inspection systems, and pattern recognition. He has been working on various research projects, some of them international, with research groups from Canada, France, Mexico and Peru. As result of his research work he has produced more than fifty scientific papers in conferences and journals and he has supervised thirteen graduate students (Master and Ph.D). He has been participating as a committee member in many national and international conferences. He also has been a reviewer for many national and international academic journals.
​
Lectures
1. Basics for 3D reconstruction
- Pinole model.
- Camera calibration.
- The Fundamental matrix.
2. Structure from Motion (SFM)
- Triangulation.
- SFM from Multiple views.
- 3D reconstruction examples.
3. Procesing 3D clouds of points
- Filtering.
- Descriptors.
- Segmentation.
4. Applications in Outdoor scenes
- Classification.
- Detection of electric towers.
- Classification of vegetative structures on coffee.
Ph.D. Industrial Automation, Institut National des Sciences Appliquées at Lyon
Associate Professor, Universidad Nacional de Colombia.
Flavio Prieto
Abstract
Image motion estimate provides powerful cues for visual systems, and has been exploited in a wide variety of applications including autonomous navigation, action recognition, human-machine interaction, three-dimensional (3D) reconstruction. In this course, we will overview main theoretical aspects of 2D and 3D motion estimation from images. Moreover, we will cover motion estimation applications, state-of-the-art strategies and implementation details.
​
Bio
Since January 2015, He is an Associate Professor in the department of Electronics at the Pontificia Universidad Javeriana in Bogotá, Colombia. He completed his PhD in Computer Vision at Inria Grenoble, France, on November 2014, under the supervision of Prof. James Crowley and Frederic Devernay at the University of Grenoble. He was also a visiting researcher in the Computer Vision group at the University of Freiburg, Germany, during the autumn semester of 2013, working with Prof. Thomas Brox. He is a member of the programme committee of conferences including ICCV, ECCV, and CVPR and regularly He review for major computer vision journals including IJCV and JVCIR. From July 2007 to December 2014 He was an Assistant Professor in the department of Electronics at the Pontificia Universidad Javeriana. He was also visiting Inria Grenoble during the summer of 2009, working with Peter Sturm. Previously, He was a research assistant at the Laboratorio de Señales at the Universidad de los Andes, Bogotá, where He worked with Prof. Alfredo Restrepo. He completed his Master in Electronics and Computer Science, on July 2006, at the Universidad de los Andes. He did his Bachelor in Electronics Engineering at the Universidad Nacional de Colombia in Manizales. His major interests are Computer Vision and Pattern Recognition with special interest in motion estimation for scene understanding.
​
Lectures
1. Optical flow estimation
- Motion estimation from images.
- Local and Variational methods.
- Discrete optimization optical flow.
- Real-time optical flow.
2. Optical flow applications
- Motion segmentation.
- Action recognition.
- Non-rigid structure from motion.
3. Scene flow estimation
- Scene flow from stereo.
- Scene flow from RGBD images.
- Scene flow applications.
Ph.D. Computer Vision, University of Grenoble
Associate Professor, Pontificia Universidad Javeriana.
Julián Quiroga Sepúlveda
Juan Carlos Niebles
Ph.D. Electrical Engineering, Princeton University
Senior Research at the Stanford AI Lab and Associate Director of Research at the Stanford-Toyota Center for AI Research.
Abstract 1
Future intelligent vehicles are a very appealing platform for advancing AI research. A fully autonomous vehicle will require a deep and accurate understanding of the environment under diverse and challenging conditions, advanced reasoning capabilities and robust planning under uncertainty. Artificial Intelligence will be the core technology that enables these features. In this talk, I will give an overview of the recent advances in Computer Vision and Artificial Intelligence developed at the Stanford AI Laboratory that tackle some of these issues.
​
Abstract 2
Humans are probably the most important subject in the many hours of video that are recorded and consumed every minute. Computer vision technology for automatic recognition of human activities and actions has the potential to enable many applications by understanding the semantics of events and activities depicted in such videos. In this talk, I'll give an overview of our work towards the next generation of activity understanding algorithms that are capable of recognizing a large number of activities, localizing them within long video sequences, parsing and describing complex events and even anticipating and predicting actions before they occur.
​
Bio
Juan Carlos Niebles received an Engineering degree in Electronics from Universidad del Norte (Colombia) in 2002, an M.Sc. degree in Electrical and Computer Engineering from University of Illinois at Urbana-Champaign in 2007, and a Ph.D. degree in Electrical Engineering from Princeton University in 2011. He is a Senior Research Scientist at the Stanford AI Lab and Associate Director of Research at the Stanford-Toyota Center for AI Research since 2015. He is also an Assistant Professor of Electrical and Electronic Engineering in Universidad del Norte (Colombia) since 2011. His research interests are in computer vision and machine learning, with a focus on visual recognition and understanding of human actions and activities, objects, scenes, and events. He is a recipient of a Google Faculty Research award (2015), the Microsoft Research Faculty Fellowship (2012), a Google Research award (2011) and a Fulbright Fellowship (2005).
​
Lectures
1. Frontiers of Computer Vision and Artificial Intelligence for Smart Vehicles.
2. Towards Detailed Visual Understanding of Human Activities.
Abstract
This short course will present recent developments on deep learning that have brought significant improvements on visual recognition tasks. The lectures will be divided into two modules: Convolutional Neural Networks for image classification and Fully Convolutional Networks for semantic segmentation. Each module will have both a theoretical and a practical components. The theory will cover problem definition, technical aspects of the proposed solution and application on the respective computer vision benchmarks. Complementarily, each module will present hands-on examples with recent neural network architectures.
​
Bio
He received a Ph.D. with honors in Applied Mathematics from the Université Paris-Dauphine in 2005. He was a Research Scientist with the Computer Vision group at UC Berkeley from 2007 to 2014. He currently holds a faculty position at Universidad de los Andes in Colombia. His research interests are in computer vision, where he has worked on a number of problems, including perceptual grouping, object recognition and the analysis of biomedical images.
​
Lectures
1. Image Classification
2. Semantic Segmentation
Ph.D. Applied Mathematics, Universite Paris-Dauphine
Assistant Professor, Universidad de los Andes
Pablo Arbelaez
