Revamping FAPESP CV

fapesp-cv.tex

@@ -3,7 +3,7 @@
 \typeout{ Andre Anjos <[email protected]>                                  }
 \typeout{ ====================================================================}
 
-\documentclass[11pt,a4paper,sans]{moderncv}
+\documentclass[10pt,a4paper,sans]{moderncv}
 \usepackage[T1]{fontenc}
 
 % styles: 'casual' (default), 'classic', 'oldstyle','banking'
@@ -13,11 +13,10 @@
 \moderncvcolor{blue}
 
 % sorting=none will preserve the bib file order
-\usepackage{csquotes}
 \usepackage[backend=biber,maxbibnames=99,bibstyle=publist,plauthorhandling=highlight,marginyear=false,sorting=ddnt]{biblatex}
+%\usepackage[style=verbose-ibid,backend=biber,sorting=none,maxbibnames=3]{biblatex}
 \setlength\bibitemsep{1.5\itemsep}
 \plauthorname[André]{Anjos}
-
 \addbibresource{fapesp.bib}
 
 % adjust the page margins
@@ -74,33 +73,6 @@
 
 \section{André Anjos -- Researcher, Head of Biosignal Processing Group}
 
-André Anjos received his Ph.D. degree in signal processing from the
-\href{https://www.ufrj.br/}{Federal University of Rio de Janeiro} in 2006. He
-joined the \href{https://atlas.ch/}{ATLAS Experiment} at European Centre for
-Particle Physics (\href{https://www.cern.ch/}{CERN}, Switzerland) from 2001
-until 2010 where he worked in the development and deployment of the Trigger and
-Data Acquisition systems that are nowadays powering the discovery of the Higgs
-boson. During his time at \href{https://www.cern.ch/}{CERN}, André studied the
-application of neural networks and statistical methods for particle recognition
-at the trigger level and developed several software components still in use
-today. In 2010, André joined the
-\href{https://www.idiap.ch/en/scientific-research/biometrics-security-and-privacy}{Biometrics Security and Privacy Group} at the
-\href{https://www.idiap.ch/en/scientific-research/biosignal-processing}{Idiap
-Research Institute} where he worked with face and vein biometrics, presentation
-attack detection, and reproducibility in research.  Since 2018 André heads the
-\href{https://www.idiap.ch/en/scientific-research/biosignal-processing}{Biosignal
-Processing Group} at Idiap. His current research interests include medical
-applications, biometrics, image and signal processing, machine learning,
-research reproducibility and open science.  Among André's open-source
-contributions, one can cite \href{https://www.idiap.ch/software/bob}{Bob} and
-the \href{https://www.idiap.ch/software/beat}{BEAT framework} for evaluation and
-testing of machine learning systems. He teaches graduate-level
-\href{http://edu.epfl.ch/coursebook/en/fundamentals-in-statistical-pattern-recognition-EE-612}{machine learning
-courses} at the École Polytechnique Fédérale de Lausanne
-(\href{https://www.epfl.ch/}{EPFL}) and master courses at Idiap's
-\href{https://master-ai.ch/}{Master of AI}. He serves as reviewer for various
-scientific journals in pattern recognition, machine learning, and image.
-
 \input{education}
 
 \section{Professional experience}
@@ -114,180 +86,38 @@ \section{Professional experience}
 \cventry{2004--2010}{Post-doctoral Researcher}{}{University of Wisconsin, Madison, USA}{}{Development and construction of the ATLAS Trigger and Data-Acquisition Systems, at \href{http://www.cern.ch}{CERN}, Switzerland.}
 
 \defbibnote{myprenote}{Select publications which are relevant to the proposal
-or significant to my research career.  For a full list of publications, please
+    or significant to my research career.  For a full list of publications, please
 consult \url{https://anjos.ai/publications/}.}
-\renewcommand{\refname}{Relevant Publications}
+\renewcommand{\refname}{Most Significative Publications}
 \nocite{*}
 \printbibliography[prenote=myprenote]
 
-\section{Current Research Grants}
+\section{Current and Past Research Grants}
 
+\cventry{2021--2022}{TheArk "SECURE"}{}{My role: PI}{}{Budget: 330'838 CHF (148'104 CHF for Idiap)}
 
 \cventry{2019--2022}{EU H2020 "AI4EU"}{PI: Patrick Gatelier (Thales SA, FR)}{My role: Partner}{}{Budget: 20'000'000 EUR (419'621 CHF for Idiap)}
 
 \cventry{2019--2022}{EU CHIST-ERA "LEARN-REAL"}{PI: Sylvain Calinon (Idiap, CH)}{My role: Co-PI}{}{Budget: 775'837 CHF (250'000 CHF for Idiap)}
 
 \cventry{2018--2021}{EU CHIST-ERA "ALLIES"}{PI: Anthony Larcher (UNIMANS, FR)}{My role: Partner}{}{Budget: 496'620CHF (496'620 CHF for Idiap)}
 
-\cventry{2021--2022}{TheArk "SECURE"}{}{My role: PI}{}{Budget: 330'838 CHF (148'104 CHF for Idiap)}
-
-\section{Current Student Supervisions}
-
-\cventry{2021--2022}{Antonio Morais}{Master thesis: \textit{A Bayesian Approach
-to Confidence Interval applied to Medical Segmentation}}{}{}{}
-
-\cventry{2021--2022}{Driss Khalil}{Master thesis: \textit{Multi-Task Learning
-for Ophtalmology}}{}{}{}
-
-\cventry{2020--2021}{Geoffrey Raposo}{Master thesis: \textit{Active tuberculosis exclusion from frontal chest X-ray images}}{}{}{}
-
 \section{Academic Indicators}
 
-\cventry{1}{Peer-Reviewed Journals}{29}{}{}{}
-\cventry{2}{Peer-Reviewed Conferences}{59}{}{}{}
+\cventry{1}{Peer-Reviewed Journals}{30}{}{}{}
+\cventry{2}{Peer-Reviewed Conferences}{60}{}{}{}
 \cventry{3}{Book Chapters}{9}{}{}{}
-\cventry{4}{Master Thesis}{4}{}{}{}
-\cventry{5}{Ph.D Thesis}{1}{}{}{}
-\cventry{6}{Citations}{+20000 (h-index: 26)}{}{}{}
-\cventry{7}{Patents}{3}{}{}{}
+\cventry{4}{Master Thesis}{Defended: 4}{Ongoing: 4}{}{}
+\cventry{5}{Ph.D Thesis}{Defended: 1}{Ongoing: 1}{}{}
+\cventry{6}{Postdocs}{Finished: 2}{Ongoing: 1}{}{}
+\cventry{7}{Citations}{+21000 (h-index: 27)}{}{}{}
+\cventry{8}{Patents}{3}{}{}{}
 
 \section{Links}
 
 \cventry{}{Google Scholar}{\url{https://scholar.google.com/citations?user=pAfLhMoAAAAJ}}{}{}{}{}
 \cventry{}{ORCID}{\url{https://orcid.org/0000-0001-7248-4014}}{}{}{}{}
 
-\section{Other information: Overview of Recent and Impactful Contributions}
-
-\subsection{Semantic Segmentation for Medical Imaging}
-
-Since the introduction of U-Nets in 2015, the field of medical image
-segmentation has seen renewed interest bringing in a variety of fully
-convolutional (deep) neural network (FCN) architectures for binary and
-multi-class segmentation problems promising very attractive results, with
-applications in computed tomography, retinography, and histopathology to cite a
-few.  Despite the incredible progress, the lack of annotated images (due to
-cost), and rigor in the comparison of trained models has led the community to
-believe larger and more dense networks provide better results.  This is
-particularly noticeable in ophtalmological images such as those from
-bi-dimensional eye fundus photography (retinography).  While retinography is
-not used for precision diagnosis, it remains cheap and very effective means for
-mass screening.  Semantical segmentation of eye fundus structures plays a key
-role in this process.
-
-I tried to address these gaps in two different ways.  The
-first~\cite{arxiv-2019} was to conduct and publish rigorous (open source,
-reproducible) benchmarks with popular retinography datasets and
-state-of-the-art FCN models in which we: i) showed that simple transformation
-techniques like rescaling, padding and cropping of combined lower-resolution
-source datasets to the resolution and spatial composition of a
-higher-resolution target dataset can be a surprisingly effective way to improve
-segmentation quality in unseen conditions; ii) we proposed a set of plots and
-metrics that give additional insights into model performance and demonstrated
-via tables and plots how to take advantage of that information, throwing a new
-light over some published benchmarks.  We argue the performance of many
-contributions available in literature is actually quite comparable within
-standard deviation margins of each other, in spite of huge differences in the
-number of parameters for different architectures.  Finally, we made our
-findings reproducible, distributing code and documentation for future
-researchers to build upon, in the hopes to inspire future work in the
-field (\url{https://gitlab.idiap.ch/bob/bob.ip.binseg}).
-
-In a second contribution~\cite{arxiv-2020} we propose that a minimalistic
-version of a standard U-Net with 3 orders of magnitude less parameters,
-carefully trained and rigorously evaluated, closely approximates the
-state-of-the-art performance in vessel segmentation for retinography.  In
-addition, we propose a simple extension, dubbed W-Net, by concatenating two
-U-Nets together, which reaches outstanding performance on several popular
-datasets, still using orders of magnitude less learnable weights than any
-previously published approach.  This work also provide a very comprehensive
-intra and cross-dataset performance analysis, involving up to 10 different
-databases, including artery/vein multi-class semantic segmentation.
-
-
-\subsection{Contributions to Computer Vision and Deep Learning in Biometrics}
-
-I have actively worked in computer vision and deep learning (mostly) associated
-to biometric recognition, with potential application to various other tasks.
-Contributions range from the collection of datasets, the exploration of
-different methods to address and assess biometric recognition vulnerabilities,
-domain adaptation, and remote photoplethysmography.
-
-While I have worked in various contributions, I highlight here key publications
-from the past 5 years that may related to the proposal.
-
-\begin{itemize}
-
-    \item Domain Specific Units (Adaptation): In~\cite{tifs-2019}, we apply
-        domain adaptation via dedicated Domain-Specific Units (DSU), with an
-        application to Heterogeneous Face Recognition.  In this class of
-        problems, one wants to recognize an individual across different
-        spectral data,  based on the representation on a principal spectrum
-        (e.g., visual).  It is a challenging task because multi-spectral data
-        for covering large populations is rare, which in turn stymies the
-        training of deep convolution-based architectures for this task.  We
-        developed a mechanism to adapt the parameters of models pre-trained on
-        large visual spectral face recognition datasets, which are readily
-        available.  My contributions are directly related to core idea of this
-        work.
-
-  \item Anomaly Detection and Robustness: Biometric recognition systems are
-      exposed to presentation attacks, and dectectors (PAD) for this purpose
-      are required building blocks of thrustworthy systems.  Most PAD systems
-      work discriminatively, trying to separate attacks from \textit{bona fide}
-      presentations.  We showed this technique does not generalize well to
-      unseen presentation attacks.  We explored, for the first time, alternate
-      approaches by joint-modelling client identity as a way to calibrate PAD
-      output scores~\cite{tifs-2015}, showing increased robustness to unseen
-      attacks.  More recently, we also showed that solely modelling
-      \textit{bona fide} presentations is also an effective way to increased
-      PAD robustness~\cite{icb-2018}.  Finally, we showed that by adding
-      heterogenous inputs to PAD systems can improve their
-      robustness~\cite{tifs-2019-2} to achieve state-of-the-art performance,
-      even to unseen conditions during training.
-
-\end{itemize}
-
-This work was published as book chapters, international peer-reviewed
-scientific journals (including articles at very high-impact factor journals)
-and in peer-reviewed conference papers totalling a few thousand citations.
-Accompanying software packages for suh contributions were released publicly,
-under an open-source license. For details and more links, please refer to the
-applicant's Research Output.
-
-\subsection{Reproducible Research}
-
-We've been actively looking at the reproducibility of published work and how to
-lower the entrance barrier of publication readers, converting them into engaged
-users of methods we create.  We argue it is insufficient, in most cases, to
-only publish software leading to results if original data remains inaccessible.
-In particular~\cite{acmmm-2012}, we note that reproducibility should imply
-in the following characteristics: repeatability, shareability, extensibility
-and stability, which is not guaranteed by most published material to date.  We
-proposed an open-source software suite called Bob
-(\url{https://www.idiap.ch/software/bob}) that possesses such characteristics,
-demonstrating its flexibility to various tasks including Medical Image
-Segmentation (\url{https://gitlab.idiap.ch/bob/bob.ip.binseg}), Remote
-Photoplethysmography (\url{https://gitlab.idiap.ch/bob/bob.rppg.base}), and
-Biometric Person Recognition (\url{https://gitlab.idiap.ch/bob/bob.bio.base}).
-
-From another perspective, there are legitimate cases in which raw data leading
-to research conclusions cannot be published.  Furthermore, in a growing number
-of use-cases, the availability of both software does not translate to an
-accessible reproducibility scenario.  The user, for example, may not have the
-necessary equipment to perform the analysis.  To bridge this gap, we built an
-open platform for research (\url{https://www.beat-eu.org/platform}) in
-computational sciences related to pattern recognition and machine learning, to
-help on the development, reproducibility and certification of results obtained
-in the field~\cite{icml-2017-1}.  The BEAT platform is distributed under an
-open-source license (\url{https://www.idiap.ch/software/beat/}).
-
-Both projects are still active and support past and future work at Idiap and
-beyond.  We conducted (and will continue doing) lectures to both master and
-graduate students about reproducibility in data science (refer to the
-applicant's CV under the section "Teaching" for details).  There are currently
-$\sim$180 \textit{direct} citations to publications about Bob and BEAT core
-frameworks.
-
 \end{document}
 
 \typeout{ *************** End of file cv *************** }