Sandra Avila

Sandra Avila
University of Campinas | UNICAMP · Institute of Computing

PhD in Computer Science
Assistant Professor

About

116
Publications
167,965
Reads
How we measure 'reads'
A 'read' is counted each time someone views a publication summary (such as the title, abstract, and list of authors), clicks on a figure, or views or downloads the full-text. Learn more
2,338
Citations
Introduction
Hello, I'm an Assistant Professor and Research Scientist in the Institute of Computing at the University of Campinas (Unicamp), Brazil. I'm also a faculty member of the RECOD.ai Lab (REasoning for COmplex Data). www.ic.unicamp.br/~sandra
Additional affiliations
February 2017 - present
University of Campinas
Position
  • Professor (Assistant)
February 2017 - present
University of Campinas
Position
  • Professor (Assistant)
April 2016 - February 2017
University of Campinas
Position
  • PostDoc Position
Education
September 2009 - June 2013
Sorbonne Université
Field of study
  • Image representation; Machine learning
October 2008 - June 2013
Federal University of Minas Gerais
Field of study
  • Image representation; Machine learning
February 2007 - September 2008
Federal University of Minas Gerais
Field of study
  • Video summarization; Feature extraction

Publications

Publications (116)
Article
The fast evolution of digital video has brought many new multimedia applications and, as a consequence, has increased the amount of research into new technologies that aim at improving the effectiveness and efficiency of video acquisition, archiving, cataloging and indexing, as well as increasing the usability of stored videos. Among possible resea...
Article
Full-text available
As web technologies and social networks become part of the general public's life, the problem of automatically detecting pornography is into every parent's mind — nobody feels completely safe when their children go online. In this paper, we focus on video-pornography classification, a hard problem in which traditional methods often employ still-ima...
Article
Recent literature has explored automated pornographic detection — a bold move to replace humans in the tedious task of moderating online content. Unfortunately, on scenes with high skin exposure, such as people sunbathing and wrestling, the state of the art can have many false alarms. This paper is based on the premise that incorporating motion inf...
Conference Paper
Full-text available
The online sharing and viewing of Child Sexual Abuse Material (CSAM) are growing fast, such that human experts can no longer handle the manual inspection. However, the automatic classification of CSAM is a challenging field of research, largely due to the inaccessibility of target data that is-and should forever be-private and in sole possession of...
Preprint
Full-text available
Skin cancer is a major public health problem that could benefit from computer-aided diagnosis to reduce the burden of this common disease. Skin lesion segmentation from images is an important step toward achieving this goal. However, the presence of natural and artificial artifacts (e.g., hair and air bubbles), intrinsic factors (e.g., lesion shape...
Article
Full-text available
Citrus juices and fruits are commodities with great economic potential in the international market, but productivity losses caused by mites and other pests are still far from being a good mark. Despite the integrated pest mechanical aspect, only a few works on automatic classification have handled images with orange mite characteristics, which mean...
Article
Full-text available
Background: Optimal control of traditional risk factors only partially attenuates the exceeding cardiovascular mortality of individuals with diabetes. Employment of machine learning (ML) techniques aimed at the identification of novel features of risk prediction is a compelling target to tackle residual cardiovascular risk. The objective of this s...
Article
Full-text available
Automatically describing images using natural sentences is essential to visually impaired people’s inclusion on the Internet. This problem is known as Image Captioning. There are many datasets in the literature, but most contain only English captions, whereas datasets with captions described in other languages are scarce. We introduce the #PraCegoV...
Preprint
Full-text available
Background Optimal control of traditional risk factors only partially attenuates the exceeding cardiovascular mortality of individuals with diabetes. Employment of machine learning (ML) techniques aimed at identification of novel features of risk prediction is a compelling target to tackle residual cardiovascular risk. Objective To identify clinic...
Preprint
Full-text available
Citrus juices and fruits are commodities with great economic potential in the international market, but productivity losses caused by mites and other pests are still far from being a good mark. Despite the integrated pest mechanical aspect, only a few works on automatic classification have handled images with orange mite characteristics, which mean...
Preprint
Full-text available
This paper shows that CIDEr-D, a traditional evaluation metric for image description, does not work properly on datasets where the number of words in the sentence is significantly greater than those in the MS COCO Captions dataset. We also show that CIDEr-D has performance hampered by the lack of multiple reference sentences and high variance of se...
Chapter
Natural disaster prediction is one of the main concerns of authorities globally. Disasters cause large-scale psychological, social, and economic damage; therefore, techniques to predict such events are essential to minimize their impacts. However, despite all efforts to estimate the occurrence of a disaster, making an accurate and robust forecast i...
Conference Paper
Full-text available
Este artigo apresenta um estudo sobre vieses gerados no aprendizado de máquina e as suas implicações na sociedade — morais, éticas e sociais. Fazemos uma releitura de um framework que posiciona os diferentes tipos de vieses nas etapas do processo de aprendizado de máquina, desde o pré-processamento, passando pela coleta dos dados, até o pós-process...
Preprint
Full-text available
Self-supervised pre-training appears as an advantageous alternative to supervised pre-trained for transfer learning. By synthesizing annotations on pretext tasks, self-supervision allows to pre-train models on large amounts of pseudo-labels before fine-tuning them on the target task. In this work, we assess self-supervision for the diagnosis of ski...
Preprint
Full-text available
Despite the growing availability of high-quality public datasets, the lack of training samples is still one of the main challenges of deep-learning for skin lesion analysis. Generative Adversarial Networks (GANs) appear as an enticing alternative to alleviate the issue, by synthesizing samples indistinguishable from real images, with a plethora of...
Preprint
Full-text available
Automatically describing images using natural sentences is an important task to support visually impaired people's inclusion onto the Internet. It is still a big challenge that requires understanding the relation of the objects present in the image and their attributes and actions they are involved in. Then, visual interpretation methods are needed...
Article
Objectives Traditional risk scores improved the definition of the initial therapeutic strategy in acute coronary syndrome (ACS), but they were not designed for predicting long-term individual risks and costs. In parallel, attempts to directly predict costs from clinical variables in ACS had limited success. Thus, novel approaches to predict cardiov...
Conference Paper
Full-text available
Recent works propose the use of UNets as fault predictor in seismic panels [Wu et al., 2019]. However, empirical tests show that just pairing seismic data with fault segmentation panels (or classes of faulted angles) leads to hard to train and unstable models that, usually, are highly dependent on the statistical distribution of training data. Many...
Conference Paper
Full-text available
Having realistic synthetic data is essential to test new techniques in Geophysics. Those techniques, when applied to the seismic processing pipeline, needs to be robust and capable to deal with realistic situations. Thus, in this work, we propose the Neural Noise Enrichment (NNE) technique, to generate perceptually realistic synthetic data. This st...
Conference Paper
Full-text available
Conventional seismic-migration techniques, while extremely valuable in modern imaging workflows, produce an inherently blurred representation of Earth's reflectivity. In this context, generative adversarial networks (GANs) are capable of learning how to map blurred or noisy data into clean data while preserving critical seismic attributes. In this...
Conference Paper
Melanoma is the most lethal type of skin cancer. Early diagnosis is crucial to increase the survival rate of those patients due to the possibility of metastasis. Automated skin lesion analysis can play an essential role by reaching people that do not have access to a specialist. However, since deep learning became the state-of-the-art for skin lesi...
Conference Paper
O câncer de pele é de longe o tipo mais comum de câncer. O diagnóstico precoce é fundamental para o tratamento do paciente, melhorando significativamente as taxas de sobrevida. O aprendizado profundo tornou-se o estado da arte na análise de lesões de pele, e os dados se tornaram um fator decisivo para impulsionar as soluções. O objetivo principal d...
Conference Paper
Full-text available
Pests and diseases are relevant factors for production losses in agriculture and, therefore, promote a huge investment in the prevention and detection of its causative agents.In many countries, Integrated Pest Management is the most widely used process to prevent and mitigate the damages caused by pests and diseases in citrus crops. However, its re...
Preprint
Segmenting skin lesions images is relevant both for itself and for assisting in lesion classification, but suffers from the challenge in obtaining annotated data. In this work, we show that segmentation may improve with less data, by selecting the training samples with best inter-annotator agreement, and conditioning the ground-truth masks to remov...
Preprint
Full-text available
Data-driven models are now deployed in a plethora of real-world applications - including automated diagnosis - but models learned from data risk learning biases from that same data. When models learn spurious correlations not found in real-world situations, their deployment for critical tasks, such as medical decisions, can be catastrophic. In this...
Preprint
Full-text available
Pests and diseases are relevant factors for production losses in agriculture and, therefore, promote a huge investment in the prevention and detection of its causative agents. In many countries, Integrated Pest Management is the most widely used process to prevent and mitigate the damages caused by pests and diseases in citrus crops. However, its r...
Article
Agricultural applications such as yield prediction, precision agriculture and automated harvesting need systems able to infer the crop state from low-cost sensing devices. Proximal sensing using affordable cameras combined with computer vision has seen a promising alternative, strengthened after the advent of convolutional neural networks (CNNs) as...
Chapter
Current art for automated diabetic retinopathy (DR) detection resides on highly abstract data-driven approaches. Usually, those approaches receive an image as input and spit the response out—that might be resulting of only one network or ensembles—and are not easily explainable. In this work, we propose an accountable referable DR detector that pro...
Conference Paper
Deep Learning is a subfield of machine learning methods based on artificial neural networks. Thanks to the increased data availability and computational power, such as Graphic Process Units (GPU), training deep networks - a time-consuming process - became possible. Cloud computing is an excellent option to acquire the computational power to train t...
Article
Deep learning fostered a leap ahead in automated skin lesion analysis in the last two years. Those models, however, are expensive to train and difficult to parameterize. Objective: We investigate methodological issues for designing and evaluating deep learning models for skin lesion analysis. We explore ten choices faced by researchers: use of tran...
Preprint
Full-text available
Generative Adversarial Networks fostered a newfound interest in generative models, resulting in a swelling wave of new works that new-coming researchers may find formidable to surf. In this paper, we intend to help those researchers, by splitting that incoming wave into six "fronts": Architectural Contributions, Conditional Techniques, Normalizatio...
Conference Paper
Diffractions play a significant role in seismic processing and imaging since they can image structures smaller than the seismic wavelength, such as discontinuities, faults, and pinch-outs. The traveltime of a non-migrated stacked diffraction event typically has a hyperbolic shape around its apex, which collapses after a migration procedure. We can...
Conference Paper
Deep Learning experiments require large amounts of labeled data, but few annotated seismic datasets are available and annotation is a time-consuming, expensive activity. Synthetic modeled datasets may be a viable alternative. However, they lack the variability and intricacies of a real data signal. Moreover, methods that add colored noises are not...
Preprint
Full-text available
Agricultural applications as yield prediction, precision agriculture and automated harvesting need systems able to infer the culture state from low cost sensing devices. Proximal sensing using affordable cameras combined to computer vision have seen a promising alternative, strengthened after the advent of convolutional neural networks (CNNs) as an...
Preprint
Full-text available
Agricultural applications as yield prediction, precision agriculture and automated harvesting need systems able to infer the culture state from low cost sensing devices. Proximal sensing using affordable cameras combined to computer vision have seen a promising alternative, strengthened after the advent of convolutional neural networks (CNNs) as an...
Preprint
Full-text available
In this work, we explore the issue of the inter-annotator agreement for training and evaluating automated segmentation of skin lesions. We explore what different degrees of agreement represent, and how they affect different use cases for segmentation. We also evaluate how conditioning the ground truths using different (but very simple) algorithms m...
Conference Paper
Full-text available
Convolutional neural networks (CNNs) deliver exceptional results for computer vision, including medical image analysis. With the growing number of available architectures, picking one over another is far from obvious. Existing art suggests that, when performing transfer learning, the performance of CNN architectures on ImageNet correlates strongly...
Article
Full-text available
Automatic classification of sensitive content in remote sensing images, such as drug crop sites, is a promising task, as it can aid law-enforcement institutions in fighting illegal drug dealers worldwide, while, at the same time, it can help monitor legalized crops in countries that regulate them. However, existing art on detecting drug crops from...
Conference Paper
Full-text available
Watching cartoons can be useful for children's intellectual, social and emotional development. However, the most popular video sharing platform today provides many videos with Elsagate content. Elsagate is a phenomenon that depicts childhood characters in disturbing circumstances (e.g., gore, toilet humor, drinking urine, stealing). Even with this...
Conference Paper
Full-text available
Violence detection in videos aims to identify whether a violent action occurred within a video stream. Effective tools for intelligent video analysis are highly demanded, specially to determine violence in video streams. Such solution could have applications in detecting inappropriate behaviors in video feeds, aiding law-enforcement in forensic cas...
Preprint
Full-text available
Convolutional neural networks (CNNs) deliver exceptional results for computer vision, including medical image analysis. With the growing number of available architectures, picking one over another is far from obvious. Existing art suggests that, when performing transfer learning, the performance of CNN architectures on ImageNet correlates strongly...
Conference Paper
Full-text available
Melanoma is the deadliest form of skin cancer. Automated skin lesion analysis plays an important role for early detection. Nowadays, the ISIC Archive and the Atlas of Dermoscopy dataset are the most employed skin lesion sources to benchmark deep-learning based tools. However, all datasets contain biases, often unintentional, due to how they were ac...
Preprint
Full-text available
Melanoma is the deadliest form of skin cancer. Automated skin lesion analysis plays an important role for early detection. Nowadays, the ISIC Archive and the Atlas of Dermoscopy dataset are the most employed skin lesion sources to benchmark deep-learning based tools. However, all datasets contain biases, often unintentional, due to how they were ac...
Preprint
Full-text available
Watching cartoons can be useful for children's intellectual, social and emotional development. However, the most popular video sharing platform today provides many videos with Elsagate content. Elsagate is a phenomenon that depicts childhood characters in disturbing circumstances (e.g., gore, toilet humor, drinking urine, stealing). Even with this...
Conference Paper
This paper analyzes the cost-benefit of using EC2 instances, specif- ically the p2 and p3 virtual machine types, which have GPU accelerators, to execute a machine learning algorithm. This analysis includes the runtime of a convolutional neural network executions, and it takes into consideration the necessary time to stabilize the accuracy value wit...
Article
Prior art on automated screening of diabetic retinopathy and direct referral decision shows promising performance; yet most methods build upon complex hand-crafted features whose performance often fails to generalize. Objective: We investigate data-driven approaches that extract powerful abstract representations directly from retinal images to pr...
Article
About 815 million people in the world suffer from chronic undernourishment. Paradoxically, 1.3 billion tonnes of food is wasted each year. When food surpluses occur, the best destination — which ensures the highest value use of edible food resources — is to redistribute these for human consumption. In this vein, we propose the Combating Waste app t...
Article
Detectar precocemente o câncer de pele é crucial: a taxa de sobrevivência é muito alta — cerca de 95% — para o diagnóstico precoce, mas cai substancialmente — para 10% a 15% — se o câncer atingir seus estágios finais. Neste contexto, o principal objetivo deste trabalho é facilitar a identificação precoce das lesões, melhorando dessa forma o prognós...
Article
Despite the efficient solutions in pornography detection literature, specific solutions for sensitive content in cartoons have not been developed yet. In this work, we evaluate how state-of-the-art solutions for natural videos (with humans) perform in cartoons. Also, we propose a new method with higher accuracy, showing that treating cartoons indep...
Article
In this work, we modeled the problem of detection of fruit and leaves in viticulture for proximal applications as a supervised machine learning task. We created and manually labeled a database of images obtained at Guaspari Winery. In total, the database consists of 11.883 images of bunch of grapes and leaves. We trained a convolutional network wit...
Chapter
Full-text available
Deep learning models show remarkable results in automated skin lesion analysis. However, these models demand considerable amounts of data, while the availability of annotated skin lesion images is often limited. Data augmentation can expand the training dataset by transforming input images. In this work, we investigate the impact of 13 data augment...
Chapter
Full-text available
Skin cancer is by far the most common type of cancer. Early detection is the key to increase the chances for successful treatment significantly. Currently, Deep Neural Networks are the state-of-the-art results on automated skin cancer classification. To push the results further, we need to address the lack of annotated data, which is expensive and...
Preprint
Full-text available
Deep learning models show remarkable results in automated skin lesion analysis. However, these models demand considerable amounts of data, while the availability of annotated skin lesion images is often limited. Data augmentation can expand the training dataset by transforming input images. In this work, we investigate the impact of 13 data augment...
Technical Report
Full-text available
This extended abstract describes the participation of RECOD Titans in parts 1 to 3 of the ISIC Challenge 2018 "Skin Lesion Analysis Towards Melanoma Detection" (MICCAI 2018). Although our team has a long experience with melanoma classification and moderate experience with lesion segmentation, the ISIC Challenge 2018 was the very first time we worke...
Conference Paper
Full-text available
Detecting violence in videos through automatic means is significant for law enforcement and analysis of surveillance cameras with the intent of maintaining public safety. Moreover, it may be a great tool for protecting children from accessing inappropriate content and help parents make a better informed decision about what their kids should watch....