I am a faculty member of the RECOD Lab — my research is often performed in cooperation with my colleagues there.
Smart Services for a Changing Society
Let’s apply our research to societal needs !
I am very interested in how we, as Scientists and Engineers, can improve Education, Health and public services.
If we want societal impact, we must go beyond theoretical advances on Computer Vision, Pattern Recognition, and Machine Intelligence. We have to make our technology affordable, enjoyable, durable, easy to learn. We must take into account legality and privacy. We must think about interactions, interfaces, and ecosystems.
I am currently very interested in computer-aided diagnosis from imagery. My team and I are working to improve screening for Diabetic Retinopathy from eye-fundus images. We are also working to create automated screening for melanoma.
We are also working on Education in several levels: Computer Science outreach (in particular for girls — see below), creation and diffusion of Computer Science courseware in Portuguese, and research of technology for education.
We are currently developing affordable solutions for classroom response-systems, that obtain in real time the reactions and answers of the students. Cutting-edge active learning practices like peer-instruction or flipped classroom are easier to implement if the instructor can conveniently collect the answers of the students as the class progresses. Ordinary CRS are implemented with “clickers”, small radiotransmitters that are given to the students. Clicker-based systems are, however expensive, with a TCO that may reach tens of thousands of dollars. Because of that, alternative solutions have been proposed, like “bring your own device” (BYOD), in which the students bring their own smartphones, tablets or notebooks and the physical clicker is replaced by an application or app. Computer Vision-based solutions also appear very attractive, where the system observes the audience using a webcam, or the camera of a tablet or smartphone and paper cards with 2D barcodes are used to collect the answers.
Large-scale Machine-Learning for Multimedia
Let’s also push the envelope in Machine Learning !
Currently I’m focusing in large-scale implementations of Deep Neural Networks and Graphical Models for large scale multimedia corpora. I have a long experience of working with images and videos, but I am currently working with other types of spatial-temporal signals, like accelerometer data, gestures, system logs, etc.
I’m particularly interested in the deployment of very large-scale Machine Learning in architectures like GPU / cloud-computing / reconfigurable architectures.
I’m also interested in mobile visual search, the deployment of image/video retrieval and classification to mobile devices like cameras, smartphones, tablets and wearables. This deployment scenario poses many interesting challenges for the design of effective feature extraction, because of the limitations on processing power, energy consumption, available network bandwidth, etc., many of which vary dynamically (e.g. as the user switches from Wifi to 3G).
Teaching & Extension
Bringing Diversity to Engineering
I am committed to advancing opportunities in STEM (Science, Technology, Engineering and Math) for students and researchers who traditionally are underrepresented in those careers : women, people of color, and LGBT people.
I currently cooperate with IEEE WIE South Brazil in a number of initiatives, including hackdays for girls and women, bringing interest in STEM to high-school girls, and fostering an welcoming environment for girls in our School.
Teaching and Learning Engineering
Ensino e Aprendizagem em Engenharia
Since August 2012, I am coordinating, with my colleagues Prof. Alim Gonçalves and Prof. Renato Lopes, the EAE group (an acronym, in Portuguese, for Teaching and Learning Engineering, endearingly pronounced [i.aˈiː]).
EAE is a casual, informal, self-ruling and egalitarian group, whose aim is not advancing policies nor seeking consensus. Its mission is being a channel for the exchange of ideas and experiences, aiming at the improvement of the classroom practices of individual teachers or small groups of teachers. It is a bottom-up crusade to improve Learning and Teaching of the STEM (Science, Technology, Engineering and Math) courses in our University, especially at the undergraduate level.
From August 2012 to June 2013 the group was organized around semi-structured / semi-chaotic “brainstorming” meetings where we formed a community of teachers concerned with the problems of undergraduate teaching, and interested in seeking solutions. During this first year, we collected ideas, complains, propositions, counter-propositions — it was a time for idea gathering.
After 12 months it was clear that the brainstorm cycle was exhausted. After a brief hiatus, the group resumed the meetings in a more structured way : bringing teachers to share, in a seminar format, ideas, experiences and practices in higher education. The talks are usually not offered by pedagogy specialists (although those are welcome), but by people who are in the trenches of higher-education practice, have tried and succeed (or failed) some technique, or learned some neat technique/idea/tool and want to share with the colleagues. That format went from late 2013 until early 2015.The seminars produced have been recorded and are available at the group’s YouTube channel (in Portuguese).
The future ? We don’t know. EAE is again on hiatus since all the founders — Renato, Alim and I, are currently on official positions as Undergraduate Studies Coordinators. But EAE was always intended as a dynamical group, and to react to the temperature of its community. It’s soul lies not dead, but dormant — just waiting for the love of adventurous faculty to reawaken.
Do you dare to take the relay ?
Innovative Courseware for Embedded Systems
where does it come from ?
RECOD lab is financed by money from UNICAMP, grants obtained from public agencies (FAPESP, CNPq), and projects done in cooperation with private research agencies and industrial partners (Microsoft Research, Samsung. Motorola, LGE).
The students’ scholarships are usually requested and granted independently of the projects, either by public agencies (FAPESP, CAPES, CNPq, FAPEAM) or by donations bound to tax benefits from private parties (Samsung, through the so-called “Brazilian IT Law”). Sometimes, scholarships are bound to specific industrial projects (sometimes the cased with Samsung).
Microsoft Research, through the Microsoft Azure for Research, conceded me a grant of computational resources to explore computer-aided diagnosis during March 2014–February 2015. Amazon Web Services is currently aiding that project, in a grant that start on June 2014 and is ongoing.
Google sponsored the seminars of EAE (Learning and Teaching Engineering) for the period of October 2013–July 2014.
Intel Software Brasil is sponsoring many activities of the Women in Engineering during the year of 2014, including our hackdays and special Women Sessions in the local events. Intel Software Brasil is also sponsoring the project for creation of new courseware for Computer Architecture and Embedded Systems.