WHO. 6 staff people. Tel: / Fax: Website: vision.unipv.it

Transcription

1 It has been active in the Department of Electrical, Computer and Biomedical Engineering of the University of Pavia since the early 70s. The group s initial research activities concentrated on image enhancement and restoration techniques, with a particular focus on medical imagery. Subsequently, the group s main efforts have been devoted to more advanced image processing functions, involving scene segmentation and shape characterization; a broad background has been acquired on low-level and intermediate-level vision tasks involving grey-level statistics and structural descriptions. Since the early 80s a new stream of research has been actively followed in the field of parallel architectures for vision and image processing. The group has meanwhile developed skills in high-level image processing domains, such as the management of knowledge description and learning capabilities for vision tasks. Recently new research areas have been activated on Deep Learning in Computer Vision, Deep Reinforcement Learning, Eye Tracking, Proteomics, Social Media Analysis, Mobile Mapping, Digital Heritage, Digital Humanities. ADDRESS Computer Vision and Multimedia Lab (Laboratorio di Visione Artificiale e Multimedia) Dipartimento di Ingegneria Industriale e dell Informazione Università degli Studi di Pavia Via Ferrata Pavia, ITALY Tel: / Fax: web-vision@unipv.it - Website: vision.unipv.it 6 staff people WHO Virginio Cantoni full professor and director Luca Lombardi, Marco Porta associate professors Giuseppe Lisanti, Mauro Mosconi assistant professors Alessandra Setti technician 2 post-doc researchers Piercarlo Dondi, Mirto Musci 2 contract professors Roberto Marmo, Marco Piastra 2 Phd Students Gianluca Gerard, Haochen Wang

2 Deep Learning in Computer Vision Several applications:? Image Enhancement Face Recognition Image Colorization Pixel Inference Learn a discriminative model to recognize persons. Collect thousands images for thousands persons and train a deep neural network (data driven approach): Deep Convolutional Neural Network

3 Deep Reinforcement Learning for Collaborative Robotics Virtualization of a real-world robot Complete 3D mesh Movable parts Joints and direct kinematic chain Moving obstacle Incremental autonomous learning Learning obstacle avoidance in a virtualized simulation environment Robust avoidance strategy Target

4 Eye Tracking Explicit and Implicit Gaze-Based Communication Gaze Input Using eye tracking as an assistive technology or as an additional input channel (besides keyboard, mouse, etc.) Soft Biometrics Identifying or verifying the identity of people from the way they look at specific stimuli (e.g., faces) E-Learning Understanding learners behavior and detecting possible comprehension problems Automotive Studying the driver s performance through cheap eye tracking solutions

5 Proteomics Geometrical motif extraction in the secondary structure of proteins Protein structure define its biological behavior. We work at second level to identify recurrent motifs. Using Hough Transform we developed the Cross Motif Search algorithm with MP and MPI for parallelism. We view proteins as a cloud of segments, we expect big data. Goal: discover new motifs with an innovative geometrical method. Motif Visualizer: Open Source OpenGL GUI to improve usability Collaboration and validation with biologists.

6 Social Media Analysis Mobile Mapping Network analysis Visual object recognition Analytics and insights Social commerce Customer segmentation Malware detection Social Media Mining Sentiment analysis, opinion mining Vehicle-mounted high-accuracy mobile 3D measurement system: Road and railway sign recognition Tunnel inspection Intelligent vehicle Laser scanner Georeferenced images

7 Digital Heritage 3D scan and measurement of historical instruments Handwriting analysis of historical documents Documents Matching and attribution Author 1 Unsure Reference Letters Author 1 Not Author 1 Analysis and presentation of visible and UV induced fluorescence (UVF) images UVF images Classification and Labeling UVF classified images Acquisition Visible images Kinect interactive presentation

8 Tactile images The battle of Pavia, attack of the French LEGEND Francis I F his horse F Bonnivet G des Prez N Mont-Jehan W Sanseverino O Charles III C Imperial cavalry Band B Imperial infantry E and E Digital Humanities 3D modeling A Certosa portal bas-relief Augmented reality The city of Pavia in the Renaissance The Samaritan woman at the Jacob s well

9 An application in the area of Digital Humanities Virtual reconstruction of Pavia in the 16 th century The realized 3D reconstruction has three goals: 1. creation of an interactive virtual environment of 16 th century Pavia in which buildings, churches and other landmarks are individually represented; 2. reproduction of different views of the city so that the viewer can move around freely, touring the city as depicted in the fresco; 3. development of an app for smartphones and tablets providing access to information about the area displayed, which, in particular, compares still existing buildings to those which have been reconstructed virtually.

10 An application in the area of Digital Humanities Virtual reconstruction of Pavia in the 16 th century Videos have been shot on routes that show the main buildings modeled in 3D, so that you can walk through the city of Pavia in the Renaissance. In particular, Route 9 ( m/watch?v=psflbrgzecw &feature=youtu.be) point of view starts about 150 meters high, 500 meters South of the city, and it runs at 360 around Pavia, flying around the city. Other videos show in details specific areas and salient buildings modeled in 3D as they were in the Renaissance.