Granted Patent: Open set classifier

Samsung and RECOD have the honor to inform a GRANTED PATENT at the USPTO (USA).

The document is: “METHOD FOR MULTICLASS CLASSIFICATION IN OPEN-SET SCENARIOS AND USES THEREOF”, publication number US 14/532,580 or as it was known by the nickname of “OPEN SET CLASSIFIER”.

This patent was filed in November 04th, 2014 and allowanced at November 01st , 2018.

This application claims the priority benefit of Brazilian Patent Application No. BR 10 2014 023780-1, filed in September 25th, 2014, at INPI (BR).

“The proposed method is used for classification in open-set scenarios, wherein often it is not possible to first obtain the training data for all possible classes that may arise during the testing stage. During the test phase, test samples belonging to one of the classes used in the training phase are classified based on a ratio between similarity scores, as known correct class and test samples belonging to any other class are to be rejected and classified as unknown.”

Congratulations to all inventors from Unicamp and Samsung who worked jointly in this project:

Anderson Rocha (Coordinator)
Pedro Ribeiro Mendes Júnior
Roberto Medeiros de Souza
Ricardo da Silva Torres
Bernardo Vecchia Stein
Daniel Vatanabe Pazinato
Waldir Rodrigues de Almeida
Rafael de Oliveira Werneck
Otavio Augusto Bizetto Penatti – Samsung

Advertisements
Posted in awards, blog, extra, science | Tagged , , , , , , , , , , , , , | Leave a comment

RECOD is Top 10 on international competition for melanoma classification

A team of RECOD researchers participated for the second time on the melanoma classification task at the melanoma diagnostic challenge promoted by the International Skin Imaging Collaboration. This year the number of participants grew 6 times regarding last edition, reaching 77 participants — our team got 6th place. Competitors all around the world submitted their approaches, including private companies focused on smart skin analytics.

RECOD also participated on the tasks of lesion segmentation and lesion attribute detection. RECOD’s participation in those tasks is detailed in a technical report (submitted before the official ranking was announced).

The global results will be discussed at the upcoming Medical Image Computing & Computer Assisted Intervention (MICCAI 2018), where RECOD team will also present two papers (out of ten selected) at the ISIC Skin Image Analysis Workshop, regarding new approaches for dermatologic dataset extension: techniques for data augmentation and synthetic image generation with GANs.

We gave a preview of our synthetic image generation with GANs at the 2nd International Educational Symposium of the Melanoma World on Rio de Janeiro (August, 2018), were it got the 2nd place at the best poster awards.

The melanoma team is composed by professors Eduardo Valle and Sandra Avila, Ph.D. student Michel Fornaciali, and M.Sc. students Alceu Bissoto, Fábio Perez and Vinícius Ribeiro, all RECOD members.

Ph.D. student Michel Fornaciali at the 2nd International Educational Symposium of the Melanoma World on Rio de Janeiro (August, 2018)

Posted in awards, blog, extra, publications, science | Tagged , , , , , , , , , , | Leave a comment

DéjàVu Talks on August 2018

Graph models for machine learning: K-associated graphs and Attribute-based Decision Graphs
Prof. João Bertini (FT/Limeira) [slides] [video]

Abstract: Graph-based methods consist of a powerful form of data representation and abstraction. Among other features, they allow uncovering topological relationships, visualizing structures, representing groups of data with different formats, and providing alternative measures to characterize the data. However, the majority of traditional graph-based methods, in general, presents a high computational complexity order, which limits the scope of application of these methods. In this context, alternative approaches based on graphs, such as K-associated graphs and Attribute-based Decision Graphs, present low computational complexity and, at the same time, have the advantages of graph-based learning. This talk will address general aspects of both models, as well as some of their applications to machine learning tasks, such as supervised and semi-supervised classification, learning from data streams, imputation of missing attribute values and enhancement of the data quality for classification tasks.

When: Thursday August 16th @11am.
Where: IC Amphitheater.

This slideshow requires JavaScript.


Reverse engineering of video content for forensic analysis: model-based and data-driven approaches.
Prof. Paolo Bestagini (Politecnico di Milano, Italy) [slides] [video]

Abstract: The recent development of multimedia devices and editing tools, together with the proliferation of video sharing websites, has made the acquisition, alteration, and diffusion of video content relatively easy tasks. As a consequence, it is possible to find more and more video sequences available on the Internet, but each of them has been potentially tampered with by anyone. It is then clear that the development of tools that enable the recovery of past history of video sequences in order to prove their origin and authenticity is more than an urgent necessity. In this talk, we discuss some of the video forensic techniques that have proposed in the last few years. Specifically, we will focus on blind forensic methods exploiting either model-based or data-driven solutions.

Mini vita: Paolo Bestagini received the B.Sc. and M.Sc. in Telecommunications Engineering at the Politecnico di Milano in 2008 and 2010, respectively. Since 2011, he joined the Image and Sound Processing Group (ISPG) at the Politecnico di Milano, and he received his Ph.D. in Information Technology in 2014. Since 2016 he is Assistant Professor at the Image and Sound Processing Group, Politecnico di Milano. His research interests focus on multimedia forensics and acoustic signal processing for microphone arrays. He is an elected member of the IEEE Information Forensics and Security Technical Committee, and a co-organizer of the IEEE Signal Processing Cup 2018.

When: Friday August 17th @2pm.
Where: IC Amphitheater

This slideshow requires JavaScript.

Posted in blog, DéjàVu, Keynotes, science, talk | Tagged , , , , , , | Leave a comment

A simple screening method for patients with zika virus

A recent research article in partnership of Recodists Prof. Anderson Rocha, Prof. Sandra Avila and Luiz Navarro and researchers at the Faculty of Pharmacy led by Prof. Rodrigo Catharino published at Frontiers in Bioengineering and Biotechnology journal is getting the media attention. The article, entitled “A machine learning application based in random forest for integrating mass spectrometry-based metabolomic data: a simple screening method for patients with zika virus”, presents a powerful solution against the analysis of the Zika virus presence based on high-resolution mass spectrometry data and machine-learning prediction model.

Since both mass spectrometry and machine learning approaches are well-established and largely utilized tools within their respective fields, this combination of methods emerges as a distinct alternative for clinical applications, providing a diagnostic screening — faster and more accurate — with improved cost-effectiveness when compared to existing technologies.

Posted in blog, media, publications, science | Tagged , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , , | Leave a comment

Image Provenance Analysis at Scale

In this article, the authors present a fully automated large-scale end-to-end pipeline that starts with the step of provenance image filtering (over millions of images) and ends up with the provenance graphs. The most immediate application of provenance image filtering is forensics, where the detection of manipulated images spans traditional policing to analysis for strategic intelligence. The question of the origins of suspect images has taken a prominent role recently, with the rise of so-called “fake news” on the Internet.

A comprehensive set of experiments for each stage of the pipeline is provided, comparing the proposed solution with state-of-the-art results, employing previously published datasets. In addition, this work introduces a new dataset of real-world provenance cases from the social media site Reddit, along with baseline results.

Posted in blog, DéjàVu, publications, science | Tagged , , , , , , , , , , , , , , , , , , | Leave a comment

Leveraging deep neural networks to fight child pornography in the age of social media

Over the past two decades, the nature of child pornography in terms of generation, distribution and possession of images drastically changed, evolving from basically covert and offline exchanges of content to a massive network of contacts and data sharing. Nowadays, the internet has become not only a transmission channel but, probably, a child pornography enabling factor by itself.

The use of deep convolution neural networks for sexually exploitative imagery of children (SEIC) is challenging, since those models require large amounts of training data. To bypass that problem, in this paper, the authors proposed a data-driven solution in which: (1) transfer the network parameters and configurations trained on ImageNet to the target problem detection (1-tiered solution) and (2) perform a 2-tiered transfer learning procedure, in which knowledge is transferred from the a network trained over ImageNet to the problem of adult content detection and fine-tune the network for detecting child pornography content in an image problem.

The proposed method outperform different existing solutions and seem to represent an important step forward when dealing with child pornography content detection. The solutions are encapsulated in a sandbox virtual machine ready for deployment by experts and practitioners.

Classification error reduction (%) with respect to our 2-tiered SEIC Detector considering the metric ACC reported in the article.

Posted in blog, DéjàVu, publications, science | Tagged , , , , , , , , , , | Leave a comment

DéjàVu Post-doctoral Fellowship

The Thematic Research Project, entitled DéjàVu: Feature-Space-Time Coherence from Heterogeneous Data form Media Integrity and Interpretation of Events funded by FAPESP is seeking candidates for a 2-year (renewable) post-doctoral fellowship position to work at the Reasoning for Complex Data (RECOD) Lab.

The project relies upon collaborators from all over the globe such as the University of Campinas (Brazil), University of São Paulo in São Carlos (Brazil), Federal University of Minas Gerais (Brazil), Purdue University (USA), University of Notre Dame (USA), Politécnico di Milano (Italy), University of Siena (Italy), Nanyang Technological University, NTU (Singapore) and others.

More Information about the Project

In this project, we focus on synchronizing specific events in space and time (X- coherence), fact-checking, and mining persons, objects and contents of interest from various and heterogeneous sources including — but not limited to — the internet, social media and surveillance imagery. For that, we seek to harness information from various media sources and synchronize the multiple textual and visual information pieces around the position of an event or object as well as order them so as to allow a better understanding about what happened before, during, and shortly after the event. After automatically organizing the data and understanding the order of the facts, we can devise and deploy solutions for mining persons or objects of interest for suspect analysis/tracking, fact-checking, or even understanding the nature of the said event. With demanding and sophisticated crimes and terrorist threats becoming ever more pervasive, allied with the advent and spread of fake news, our objective is to use the developed solutions to help us answering the four most important questions in forensics regarding an event: “who,” “in what circumstances,” “why,” and “how,” thus identifying the characteristics and circumstances in which an event has taken place.

The Position

This fellowship position requires research and development in Computer Vision, Machine Learning and Visual Analytics, in collaboration with graduate students and partners. The work of the fellow will be focused on machine learning and visual analytics methods to perform X-coherence, or in other words, to find out the order of facts related to an event in space and time.

It is desirable that the candidate demonstrates domain knowledge in machine learning, visual analytics, and computer vision. However, candidates with good mathematical and programming backgrounds in any of the three areas and motivation to learn the others are equally welcomed.

The post-doctoral fellowship includes a monthly stipend of R$ 7,174.80 (about USD 2,300 and EU$ 2,000), access to the health-care system of Unicamp, and research contingency funds (15% of the annual value of the fellowship, each year). For more details, check out Fapesp’s webpage.

How to Apply

Interested candidates should email Prof. Anderson Rocha, project’s coordinator, before February 20th, 2018 with:

  • A motivation letter for the application;
  • A recommendation letter from a previous supervisor;
  • Curriculum vitae with the list of publications, education background, research track-record and experience, and copy of diplomas/degree certificate(s).

Additional Information

Eligibility Criteria Ph.D. in Computer Science or related areas (in case you have any doubt about a possible related area, drop an e-mail to the Project’s Coordinator above)

Selection process It will be based on the motivation letter for the application, recommendation letter from a previous supervisor, curriculum vitae with the list of publications, education and experience, and copy of diplomas/degree certificate(s). An interview with the finalists shall take place via Telecom.

Information About Campinas

More details

Posted in blog, DéjàVu, extra | Tagged , , , | Leave a comment

RECOD celebrates its 8th year!

This year, our lab celebrated its eighth year of success together with an end of the year party. We have achieved significant growth in the international scientific community. Also, this year two professors joined RECOD: Adin Rivera and Sandra Avila.

As part of the celebration, we launched a contest for Recod’s 8th Anniversary Logo and the winning one was Yusseli Mendez with this nice orange logo:

Congratulations to all the Researchers, students, alumni, and friends that were part of this journey during this year!

This slideshow requires JavaScript.

Posted in blog, extra | Tagged , , , , , , , | Leave a comment

Talks: LIRMM and WIFS 2017, France

Prof. Anderson Rocha is visiting the Montpellier Laboratory of Informatics, Robotics and Microelectronics (LIRMM), Montpellier, France, hosted by Prof. William Puech. Two talks were/will be given:

Talk #1: Dec. 12th, 2017

Interpretation of Events from Heterogenous Data: How to Connect Existing Pieces of Evidence in Time and Space

Abstract: In this talk, we will discuss problems associated with synchronizing specific events in space and time (X-coherence), fact-checking, and mining persons, objects and contents of interest from various and heterogeneous sources including — but not limited to — the internet, social media and surveillance imagery. For that, we seek to harness information from various media sources and synchronize the multiple textual and visual information pieces around the position of an event or object as well as order them so as to allow a better understanding about what happened before, during, and shortly after the event. After automatically organizing the data and understanding the order of the facts, we can devise and deploy solutions for mining persons or objects of interest for suspect analysis/tracking, fact-checking, or even understanding the nature of the said event. Additionally, by exploring the possible existing links among different pieces of information, we aim at further designing and developing media integrity analytics tools to hint at existing forgeries, sensitive content (e.g., violent content, child pornography), and spreading patterns of multimedia objects online. With demanding and sophisticated crimes and terrorist threats becoming ever more pervasive, allied with the advent and spread of fake news, our objective is to use the developed solutions to help us answering the four most important questions in forensics regarding an event: “who,” “in what circumstances,” “why,” and “how,” thus identifying the characteristics and circumstances in which an event has taken place.

Talk #2: Dec. 19th, 2017

Media Integrity Analytics: Beyond Digital Forensics of Single Objects

Abstract: Currently, multimedia objects can be easily created, stored, (re)-transmitted, and edited for good or bad. In this sense, there has been an increasing interest in finding the structure of temporal evolution within a set of documents and how documents are related to one another overtime. This process, also known in the literature as Multimedia Phylogeny, aims at finding the phylogeny tree(s) that best explains the creation process of a set of related documents (e.g., images/videos) and their ancestry relationships. Solutions to this problem have direct applications in forensics, security, copyright enforcement, news tracking services and other areas. In this talk, we explore solutions for reconstructing the evolutionary tree(s) associated with a set of visual documents, more specifically images and videos. This can be useful for aiding experts to track the source of child pornography image broadcasting or the chain of image and video distribution in time, being extremely useful for complex different media provenance tasks. Finally, we will also discuss how to implement such solutions in large-scale considering millions of documents at the same time.


In Rennes, on Dec. 6th, Prof. Anderson gave a talk at IEEE Intl. Workshop on Information Forensics and Security (WIFS 2017) in which he presented the paper:

Authorship Attribution for Social Media Forensics, A. Rocha, W. J. Scheirer, C. W. Forstall, T. Cavalcante, A. Theophilo, B. Shen, A. R. B. Carvalho, and E. Stamatatos, in IEEE Transactions on Information Forensics and Security, vol. 12, no. 1, pp. 5-33, Jan. 2017.

This slideshow requires JavaScript.

Posted in blog, DéjàVu, Keynotes, publications, science, talk | Tagged , , , , , , , , , | Leave a comment

RECODist has been awared with INOVA Prize 2017

We are very proud to announce that Gabriel Bertocco, one of our RECODists, has been awarded with the Unicamp INOVA Prize on Introduction to Innovation 2017. Supervised by Prof. Anderson Rocha and Fernanda Andaló, the work entitled “Automatic age range estimation on mobile devices” is the result of a partnership between RECOD and Motorola. The Unicamp INOVA Prize awarded to most innovative works in applied or basic sciences made by undergraduate students under scientific tutorial at Unicamp.

Another research work from RECOD, entitled “CrowdPet: Deep learning applied to the detection of dogs in the wild”, was also among the finalists in the technological area.

The related links are (in Portuguese):

http://www.inova.unicamp.br/noticia/premio-inova-premia-3-pesquisas-com-grande-potencial-de-beneficiar-a-sociedade/

http://www.inova.unicamp.br/noticia/conheca-os-projetos-vencedores-do-premio-inova-de-iniciacao-a-inovacao/

http://www.inova.unicamp.br/noticia/finalistas-do-premio-inova-de-iniciacao-a-inovacao-tem-projetos-avaliados-no-congresso-de-ic/

Posted in awards, blog, science | Tagged , , , , , , , , | Leave a comment