Comments Off on Visual Intelligence and Machine Perception (VIMP) group

By lambertoballan, February 5, 2019

Finally we are online! Check the website of my group at the University of Padova. We conduct research in computer vision, applied machine (deep) learning, NLP and multimedia. We aim at developing artificially intelligent systems to help computers perform visual perception and recognition tasks.

Comments Off on Under the hood of Big Data and Artificial Intelligence

By lambertoballan, November 12, 2018

I recently gave a talk at “The Law of Big Data” seminar series @UniPD (you can watch the video on YouTube).

How AI and big data are affecting our society, culture and laws? Many are claiming that data is the “new oil”, and the data-driven / machine-learning paradigm is changing how we address many different problems. Self-driving cars, robot caregivers and chatbot platforms are really happening, while they were only popular sci-fi topics until a few years ago. The aim of this lecture is to introduce the key concepts that have led to these results, to highlight what are the main challenges and open problems, thus trying to unveil what’s in the box.

Comments Off on Teaching machines to see: a quest to visual intelligence

By lambertoballan, May 8, 2018

I recently gave a talk on “Teaching machines to see: a quest to visual intelligence” at the “Math&CS Pride” open-day outreach event organized by the Dept of Mathematics “Tullio Levi-Civita”, for undergrad and high school students.

It was super fun (at least for me :), and this are the slides of the talk.

Comments Off on New class on cognitive computing, applied machine learning and vision

By lambertoballan, February 22, 2018

This semester, together with Alessandro Sperduti, we are teaching a new course on cognitive computing, applied machine learning and computer vision. This new class is motivated by the recent revolution of cloud-enabled artificial intelligence services. We will give an introductory overview of deep learning, the main fields of application of machine learning and representation learning, with a particular emphasis on visual recognition problems. The class is offered to master students in computer science and data science, but it is also open to others (e.g. undergrads in CS or students from the school of engineering). More info: course webpage @UniPD, course presentation.

Comments Off on Deep Learning, Pattern Recognition and Vision, which direction?

By lambertoballan, January 29, 2018

I have been invited to contribute to the panel “Deep Learning, Pattern Recognition and Vision, which direction?” during the CVPL day in Modena, Italy, on “Why is Deep Learning so cool?” [slides available online]. The event has been opened by Naftali Tishby with a very interesting talk about “Information Theory of Deep Learning”. A similar talk is also available on YouTube.

Comments Off on European Researchers’ Night: “Science is wonder-ful!”

By lambertoballan, September 29, 2017

This 26 and 27 September 2017, I was involved in the kickoff event for European Researchers’ Night in Brussels #MSCAnight. With thousands of visitors, most of which were children, it was a smashing success! I have presented my research on machine learning and visual perception, and talked about visual illusions and teaching computers to see.

More about the event: https://goo.gl/QNkxgY, more about my work and some nice pictures.

Comments Off on Aloha CVPR’17: I will be giving two talks on July 26

By lambertoballan, July 22, 2017

CVPR is growing exponentially. This year in the beautiful Honolulu there are approx. 5K attendees. First, I will present our paper on “Localization of JPEG double compression through multi-domain convolutional neural networks” (Amerini, Uricchio, Ballan, Caldelli) at the CVPR’17 Workshop on Media Forensics. Then, I will give a keynote talk on “Exploiting noisy web data for large-scale visual recognition” at the CVPR’17 Workshop on Visual Understanding by Learning from Web Data [slides available online].

Comments Off on New papers on image tagging and webly-supervised action recognition

By lambertoballan, July 21, 2017

Our paper “Automatic Image Annotation via Label Transfer in the Semantic Space”, by T. Uricchio, myself, L. Seidenari and A. Del Bimbo, has been accepted for publication in Pattern Recognition and is now available online. It is an extended version of our KCCA-based tag propagation model presented in our ICMR’14 paper, containing more experiments and a novel tag denoising procedure.

A few days later, our paper “Learning without Prejudice: Avoiding Bias in Webly-Supervised Action Recognition” has been also accepted for publication in Computer Vision and Image Understanding (CVIU) and is now available online. Here we present a (fully) webly-supervised model for action recognition in videos. This is a joint work with F. Tombari and C. Rupprecht from TUM (Germany).

Comments Off on I will be joining the University of Padova!

By lambertoballan, February 3, 2017

Happy to share that I will join the Math & CS Department of the University of Padova as an Assistant Professor (tenure track) of Computer Science, starting in Fall 2017.

The University of Padova is one of Europe’s oldest and most prestigious seats of learning. The University of Padova is ranked first among Italian universities according to most international rankings (ARWU, US-News) and research evaluation agencies (ANVUR) [UniPD at a glance].

Universa Universis Patavina Libertas.

Comments Off on Invited Talk: Sharing Knowledge for Large Scale Visual Recognition

By lambertoballan, November 18, 2016

In the last months I gave several times this talk about my recent work on knowledge transfer for large-scale visual recognition problems (e.g. Aquifi Inc – Palo Alto, Google Research, UC Santa Cruz, MICC, U. of Parma, U. of Catania, U. of Padova, “Ca’ Foscari” U. Venice). The key idea of our work is to transfer prior contextual knowledge to novel scenes where it is hard to collect large-scale training data [slides available online].