Empirical comparison of cross-validation and internal metrics for tuning SVM hyperparameters

Hyperparameter tuning is a mandatory step for building a support vector machine classifier. In this article, the authors study some methods based on metrics of the training set itself, and not the performance of the classifier on a different test set – the usual cross-validation approach. Then, they compare cross-validation (5-fold) with Xi-alpha, radius-margin bound, generalized approximate cross validation, maximum discrepancy and distance between two classes on 110 public binary data sets.

The authors demonstrate that cross validation is the method that resulted in the best selection of the hyper-parameters, but it is also the method with one of the highest execution time. On the other hand, distance between two classes (DBTC) is the fastest and the second best ranked method. The authors also discuss that DBTC is a reasonable alternative to cross validation when training/hyperparameter-selection times are an issue and that the loss in accuracy when using DBTC is reasonably small.


Edson Duarte, Jacques Wainer, Empirical comparison of cross-validation and internal metrics for tuning SVM hyperparameters, Pattern Recognition Letters, Volume 88, 1 March 2017, Pages 6-11, ISSN 0167-8655, http://dx.doi.org/10.1016/j.patrec.2017.01.007.

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

Video pornography detection through deep learning techniques and motion information

In this paper, the authors deal with a growing issue of our connected society: automated sensitive media (pornographic, violent, gory, etc.) filtering. A range of applications has increased societal interest on the problem, e.g., detecting inappropriate behavior via surveillance cameras; or curtailing the exchange of sexually-charged instant messages, also known as “sexting”, by minors. In addition, law enforcers may use pornography filters as a first sieve when looking for child pornography in the forensic examination of computers, or Internet content. The main application, however, remains preventing uploading or accessing undesired content for certain demographics (e.g., minors), or environments (e.g., schools, workplace).

In spite of the success of deep learning techniques in the computer vision arena, their literature on pornography detection is very scarce. In this work, the authors design and develop deep learning-based approaches to automatically extracting discriminative spatio-temporal characteristics for filtering pornographic content in videos. The evaluation of the proposed techniques shows that the association of Deep Learning with the combined use of static and motion information considerably improves pornography detection. Not only over current scientific state of the art, but also over off-the-shelf software solutions.

The contributions of this paper are three-fold:

i) A novel method for classifying pornographic videos, using convolutional neural networks along with static and motion information;

ii) A new technique for exploring the motion information contained in the MPEG motion vectors;

iii) A study of different forms of combining the static and motion information extracted from questioned videos.


Mauricio Perez, Sandra Avila, Daniel Moreira, Daniel Moraes, Vanessa Testoni, Eduardo Valle, Siome Goldenstein, Anderson Rocha, Video pornography detection through deep learning techniques and motion information, Neurocomputing, Volume 230, 22 March 2017, Pages 279-293, ISSN 0925-2312, http://dx.doi.org/10.1016/j.neucom.2016.12.017.

Posted in blog, publications, science | Tagged , , , , , , , , , , , , , , | 1 Comment

RECOD wins international competition for melanoma classification

A team of RECOD researchers won the melanoma classification task at the “Skin Lesion Analysis towards Melanoma Detection” challenge promoted by the International Skin Imaging Collaboration (ISIC).

RECOD got the third place (among 23 participants) at the skin lesion classification for two lesions (melanoma and seborrheic keratosis), and the fifth place for skin lesion segmentation. For the specific task of melanoma detection — the most important in this research area — RECOD got first place. RECOD’s participation in those tasks is detailed in a technical report (submitted before the official ranking was announced).

The results will be presented by Prof. Eduardo Valle at the upcoming International Symposium of Biomedical Imaging (ISBI 2017), where the RECOD team will also present a paper about Transfer Learning and Deep Learning for skin lesion classification.

The team was composed by professors Eduardo Valle  and Sandra Avila, post-doc researcher Lin Tzy Li, Ph.D. student Michel Fornaciali, and M.Sc. students Afonso Menegola and Julia Tavares, all RECOD members.

Prof. Eduardo Valle and Michel Fornaciali were recipients of the Google Research Awards for Latin America 2016, with a project related to the automatic screening of melanoma. More details can be found at Unicamp News (in Portuguese).

RECOD Titans Melanoma Team

From left to right: Julia Tavares, Prof. Sandra Avila, Michel Fornaciali, Prof. Eduardo Valle, Dr. Lin Tzy Li, and Afonso Menegola

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

Talk: The brave new world of open-set recognition

Being part of a series of four talks that will be given at NTU Singapore, in this fourth talk Prof. Anderson Rocha (RECOD) explored the research field of Open-set Recognition. The talk comprises four parts, each of which lasting approximately 45 minutes, totalling three hours. Parts 1 and 2 were delivered in Day #1 (March, 6th) while the others will be delivered in Day #2 (March, 9th).

Abstract: Coinciding with the rise of large-scale statistical learning within the visual computing, forensics and security areas, there has been a dramatic improvement in methods for automated image recognition in myriad of applications ranging from, categorization, object detection, forensics, and human biometrics, among many others. Despite this progress, a tremendous gap exists between the performance of automated methods in the laboratory and the performance of those same methods in the field. A major contributing factor to this is the way in which machine learning algorithms are typically evaluated: without the expectation that a class unknown to the algorithm at training time will be experienced at test time during operational deployment.

The purpose of this talk is to introduce the audience to this difficult problem in statistical learning specifically in the context of visual computing, information forensics and security applications. Examples considering other areas will also be given for completeness. A number of different topics will be explored, including supervised machine learning, probabilistic models, kernel machines, the statistical extreme value theory, and case studies for applications related to the analysis of images.

Outline

Part 1: An introduction to the open set recognition problem

  • General introduction: where do we find open set problems in visual computing, information forensics and security?
  • Decision models in machine learning
  • Theoretical background: the risk of the unknown
  • The compact abating probability model (Scheirer et al., T-PAMI 2014)
  • The Open-set Nearest Neighbors classifier (Mendes Jr. et al., Machine Learning, 2017)

10-minute break

Part 2: Algorithms that minimize the risk of the unknown

  • Kernel Density Estimation
  • 1-Class Support Vector Machines (SVMs)
  • Support Vector Data Description
  • 1-vs-Set Machine (Scheirer et al. T-PAMI 2013)
  • PI-SVM (Jain et al. ECCV 2014)
  • W-SVM (Scheirer et al. T-PAMI 2014)
  • Decision Boundary Carving (Costa et al. 2014)
  • Open-set Nearest Neighbors (Mendes Jr. et al. 2017)

Part 3: Case studies related to visual computing and other areas

  • Image Classification/Recognition
  • Visual Information Retrieval
  • Detection problems (e.g., pedestrian, objects)
  • Face Recognition
  • Scene Analysis for Surveillance
  • Source Camera Attribution
  • Authorship Attribution

10-minute break

Part 4: Research opportunities and trends

  • The open set recognition problem and new feature characterization methods (e.g., deep learning)
  • Integrating open set solutions with the image characterization process directly (strongly generalizable image characterization)
  • Opportunities for novelty detection and automatic addition of classes (online adaptation)
  • Bringing the user into the loop (relevance feedback)
  • Final considerations

If you are interested in the talk’s content, the complete set of slides is available here.

Time: 2.00pm – 3.30 pm (Seated by 1.50pm)

Venue: Demo Room, ROSE Lab, Research Techno Plaza (RTP), Level 4, Border X Block, 50 Nanyang Drive, 637553

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

New research grants

We are proud to announce the following new research grants obtained recently by RECOD team:

  • (2016 – 2017) Leveraging Data-Driven Techniques for Screening of Diabetic Retinopathy and Alzheimer’s Disease. Coordinator: Prof. Dr. Anderson Rocha. Microsoft Research Azure for Research.
  • (2017 – 2018) Fighting Diabetic Retinopathy and Alzheimer’s Disease in the 21st Century: Leveraging Data-Driven Techniques for More Principled Decisions. Coordinator: Prof. Dr. Anderson Rocha. NVidia Corporation.
  • (2017 – 2020) Statistiques Robustes pour l’ApprentIssage L ́eger (Robust Statistics-based Light Learning). Associate Researcher: Prof. Dr. Anderson Rocha. General Coordinator: Prof. William Puech, University of Montpellier, France. Agence Nationale de la Recherche (ANR), France.
Posted in blog, extra | Tagged , , , , , , | Leave a comment

Talk: Sensitive Video Analysis

Being part of a series of four talks that will be given at NTU Singapore, in this third talk Prof. Anderson Rocha (RECOD) explored the research field of Sensitive Video.

Abstract: Sensitive video can be defined as any motion picture that may pose threats to its audience. Typical representatives include pornography, violence, child abuse, cruelty to animals, etc. Sensitive-content analysis represents a major concern to law enforcers, companies, tutors, and parents, due to the potential harm of such contents over minors, students, workers, etc. In this talk, we will discuss how to tackle this problem in two ways. In the first one, we aim at deciding whether or not a video stream presents sensitive content, which we refer to as sensitive-video classification. In the second one, we aim at finding the exact moments a stream starts and ends displaying sensitive content, at frame level, which we refer to as sensitive-content localization. For both cases, we design and develop effective and efficient methods, with low-memory footprint and suitable for deployment on mobile devices. During the talk, each of these methods will be explained in details. We start with a novel Bag-of-Visual-Words-based pipeline for efficient time-aware sensitive-video classification. Then we move to a novel high-level multimodal fusion pipeline for sensitive-content localization. Finally, we introduce a novel space-temporal video interest point detector and video content descriptor.

Date: 27 Feb 2017, Monday

Time: 2.00pm – 3.00 pm (Seated by 1.50pm)

Venue:  Demo Room, ROSE Lab, Research Techno Plaza (RTP), Level 4, Border X Block, 50 Nanyang Drive, 637553

ntusensitive-videoanalysis

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

RECOD Fourth Report 2009–2017

We are proud to share our latest achievement laboratory’s report, which translates into numbers the excellent efforts of a great research team at UNICAMP that counts upon six permanent professors, post-doctoral researchers, world-class collaborators all around the globe, and a large team of students in computer vision, information retrieval, machine learning and digital forensics. Congrats everyone in the team and thanks for the great dedication and hard work!

Summary

From 2009 to January, 2017 RECOD has achieved:

recodinnumbers17-00247 awards and distinctions;
141 articles in international journals (20+ a year);
223 articles in conferences (almost 30 a year);
10 book chapters;
8 patents;
14 completed post-doc fellowships;
27 graduated Ph.D. students  (~4 a year);
68 graduated M.Sc. students (~9 a year).

We have ongoing:

04 associated postdoc fellows;
27 Ph.D. students;
21 MSc. students;
03 B.Sc. students (undergraduate research — “Introduction to Science”).
 

Since its inception, the impact of the lab. (measured by Google Scholar citations) is growing fast. In 2016, we got ~1,561 citations, and in the previous year, 2015, we got ~1,461 citations. In 2014, we got ~1,300 citations. Although there are many redundant citations due to the internal cooperations within the lab., the number is still impressive. In 2013, we’ve got collectively 1,166 citations, and in 2012, 950 citations. In 2011, this number was 630, and in 2010, 550.

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

Talk: Leveraging Spatial, Spectral and Temporal Features for Spoofing Detection in Images and Videos

Being part of a series of four talks that will be given at NTU Singapore, in this second talk Prof. Anderson Rocha (RECOD) will explore the research field of Spoofing Detection in Images and Videos.

Abstract: Recent advances on biometrics, information forensics, and security have improved the recognition effectiveness of facial biometric systems. However, an ever-growing challenge is the vulnerability of such systems to spoofing attacks, in which impostor users create synthetic samples from original biometric information of a valid user and show them to the biometric system seeking to authenticate themselves as valid users. Depending on the trait used by the biometric systems, the type of attack varies with the type of material used to build the synthetic samples. For instance, in facial biometric systems, an attempted attack is characterized by the type of material the impostor uses such as a photograph, a digital video, or a 3-D mask with the facial information of a target user. In iris-based biometrics, spoofing attempts can be performed with printed photo or with contact lenses containing the iris patterns of a target user. Finally, fingerprint biometric systems can be spoofed by an impostor user in possession of replicas of the fingerprint patterns built with materials such as latex, play-doh, silicone, among other. In this talk, we will present anti-spoofing solutions whose objective is to detect attempted attacks considering these different types of attacks, in each modality. The lines of investigation include devising and developing visual representations based on spatial, temporal and spectral information and deep learning techniques.

If you are interested in the talk’s content, the slides are available here.

Date: 22 Feb 2017, Wednesday

Time: 2.00pm – 3.00 pm (Seated by 1.50pm)

Venue: Demo Room, ROSE Lab, Research Techno Plaza (RTP), Level 4, Border X Block, 50 Nanyang Drive, 637553

image-uploaded-from-ios-2

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

A graph-based ranked-list model for unsupervised distance learning on shape retrieval

Overview of the proposed Ranked-List Graph model

Overview of the proposed Ranked-List Graph model

Several re-ranking algorithms have been proposed recently. Some effective approaches are based on complex graph-based diffusion processes, which usually are time consuming and therefore inappro- priate for real-world large scale shape collections.

In this article, the authors present a novel rank-based algorithm for improving the effectiveness of shape retrieval tasks. The algorithm models each ranked list as a graph, establishing similarity connections among all top-k images. Next, a graph fusion approach is employed for obtaining a single graph representing the whole collection and exploiting the relationships encoded in the dataset manifold. Based on the fused graph, a new distance is learned and a new set of ranked lists is computed. The effectiveness of the proposed approach is demonstrated by the performance of an extensive experimental protocol considering widely used shape collections.


Daniel Carlos Guimarães Pedronette, Jurandy Almeida, Ricardo da S. Torres, A graph-based ranked-list model for unsupervised distance learning on shape retrieval, Pattern Recognition Letters, Volume 83, Part 3, 1 November 2016, Pages 357-367, ISSN 0167-8655, http://dx.doi.org/10.1016/j.patrec.2016.05.021.

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

A correlation graph approach for unsupervised manifold learning in image retrieval tasks

In this paper, the authors Daniel Pedronette and Ricardo Torres, discuss a novel unsupervised manifold learning algorithm, which aims at imitating the human behavior in judging similarity among images. The proposed algorithm exploits unlabeled contextual information encoded in the dataset manifold through the Correlation Graph for improving the effectiveness of distance/similarity measures. In this sense, the context can be seen as any complementary information about similarity among images, as the set of images in a strongly connected component.

A large set of experiments was conducted for assessing the effectiveness of the proposed approach, considering different descriptors and datasets. The high effectiveness of the manifold learning algorithm is demonstrated by the experimental results obtained in several image retrieval tasks. The effective- ness gains associated with the low computational efforts required represent a significant advantage of the discussed method when compared with existing approaches proposed in the literature.


Daniel Carlos Guimarães Pedronette, Ricardo da S. Torres, A correlation graph approach for unsupervised manifold learning in image retrieval tasks, Neurocomputing, Volume 208, 5 October 2016, Pages 66-79, ISSN 0925-2312, http://dx.doi.org/10.1016/j.neucom.2016.03.081.

Distance correlation analysis for computing the adjacency of the Correlation Graph

Distance correlation analysis for computing the adjacency of the Correlation Graph

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