RESUMEN

Producing or sharing Child Sexual Exploitation Material (CSEM) is a severe crime that Law Enforcement Agencies (LEAs) fight daily. When the LEA seizes a computer from a potential producer or consumer of the CSEM, it analyzes the storage devices of the suspect looking for evidence. Manual inspection of CSEM is time-consuming given the limited time available for Spanish police to use a search warrant. Our approach to speeding up the identification of CSEM-related files is to analyze only the file names and their absolute paths rather than their content. The main challenge lies in handling short and sparse texts that are deliberately distorted by file owners using obfuscated words and user-defined naming patterns. We present two approaches to CSEM identification. The first employs two independent classifiers, one for the file name and the other for the file path, and their outputs are then combined. Conversely, the second approach uses only the file name classifier to iterate over an absolute path. Both operate at the character n-gram level, whereas novel binary and orthographic features are presented to enrich the text representation. We benchmarked six classification models based on machine learning and convolutional neural networks. The proposed classifier has an F1 score of 0.988, which can be a promising tool for LEAs.

Asunto(s)

RESUMEN

Botnets are one of the most harmful cyberthreats, that can perform many types of cyberattacks and cause billionaire losses to the global economy. Nowadays, vast amounts of network traffic are generated every second, hence manual analysis is impossible. To be effective, automatic botnet detection should be done as fast as possible, but carrying this out is difficult in large bandwidths. To handle this problem, we propose an approach that is capable of carrying out an ultra-fast network analysis (i.e. on windows of one second), without a significant loss in the F1-score. We compared our model with other three literature proposals, and achieved the best performance: an F1 score of 0.926 with a processing time of 0.007 ms per sample. We also assessed the robustness of our model on saturated networks and on large bandwidths. In particular, our model is capable of working on networks with a saturation of 10% of packet loss, and we estimated the number of CPU cores needed to analyze traffic on three bandwidth sizes. Our results suggest that using commercial-grade cores of 2.4 GHz, our approach would only need four cores for bandwidths of 100 Mbps and 1 Gbps, and 19 cores on 10 Gbps networks.

RESUMEN

Retrieving text embedded within images is a challenging task in real-world settings. Multiple problems such as low-resolution and the orientation of the text can hinder the extraction of information. These problems are common in environments such as Tor Darknet and Child Sexual Abuse images, where text extraction is crucial in the prevention of illegal activities. In this work, we evaluate eight text recognizers and, to increase the performance of text transcription, we combine these recognizers with rectification networks and super-resolution algorithms. We test our approach on four state-of-the-art and two custom datasets (TOICO-1K and Child Sexual Abuse (CSA)-text, based on text retrieved from Tor Darknet and Child Sexual Exploitation Material, respectively). We obtained a 0.3170 score of correctly recognized words in the TOICO-1K dataset when we combined Deep Convolutional Neural Networks (CNN) and rectification-based recognizers. For the CSA-text dataset, applying resolution enhancements achieved a final score of 0.6960. The highest performance increase was achieved on the ICDAR 2015 dataset, with an improvement of 4.83% when combining the MORAN recognizer and the Residual Dense resolution approach. We conclude that rectification outperforms super-resolution when applied separately, while their combination achieves the best average improvements in the chosen datasets.

RESUMEN

Face recognition is a valuable forensic tool for criminal investigators since it certainly helps in identifying individuals in scenarios of criminal activity like fugitives or child sexual abuse. It is, however, a very challenging task as it must be able to handle low-quality images of real world settings and fulfill real time requirements. Deep learning approaches for face detection have proven to be very successful but they require large computation power and processing time. In this work, we evaluate the speed-accuracy tradeoff of three popular deep-learning-based face detectors on the WIDER Face and UFDD data sets in several CPUs and GPUs. We also develop a regression model capable to estimate the performance, both in terms of processing time and accuracy. We expect this to become a very useful tool for the end user in forensic laboratories in order to estimate the performance for different face detection options. Experimental results showed that the best speed-accuracy tradeoff is achieved with images resized to 50% of the original size in GPUs and images resized to 25% of the original size in CPUs. Moreover, performance can be estimated using multiple linear regression models with a Mean Absolute Error (MAE) of 0.113, which is very promising for the forensic field.

Asunto(s)

RESUMEN

The advantages of automatically recognition of fundamental tissues using computer vision techniques are well known, but one of its main limitations is that sometimes it is not possible to classify correctly an image because the visual information is insufficient or the descriptors extracted are not discriminative enough. An Ontology could solve in part this problem, because it gathers and makes possible to use the specific knowledge that allows detecting clear mistakes in the classification, occasionally simply by pointing out impossible configurations in that domain. One of the main contributions of this work is that we used a Histological Ontology to correct, and therefore improve the classification of histological images. First, we described small regions of images, denoted as blocks, using Local Binary Pattern (LBP) based descriptors and we determined which tissue of the cardiovascular system was present using a cascade Support Vector Machine (SVM). Later, we built Resource Description Framework (RDF) triples for the occurrences of each discriminant class. Based on that, we used a Histological Ontology to correct, among others, "not possible" situations, improving in this way the global accuracy in the blocks first and in tissues classification later. For the experimental validation, we used a set of 6000 blocks of [Formula: see text] pixels, obtaining F-Scores between 0.769 and 0.886. Thus, there is an improvement between 0.003 and [Formula: see text] when compared to the approach without the histological ontology. The methodology improves the automatic classification of histological images using a histological ontology. Another advantage of our proposal is that using the Ontology, we were capable of recognising the epithelial tissue, previously not detected by any of the computer vision methods used, including a CNN proposal called HistoResNet evaluated in the experiments. Finally, we also have created and made publicly available a dataset consisting of 6000 blocks of labelled histological tissues.

Asunto(s)

RESUMEN

This paper presents a new texture descriptor booster, Complete Local Oriented Statistical Information Booster (CLOSIB), based on statistical information of the image. Our proposal uses the statistical information of the texture provided by the image gray-levels differences to increase the discriminative capability of Local Binary Patterns (LBP)-based and other texture descriptors. We demonstrated that Half-CLOSIB and M-CLOSIB versions are more efficient and precise than the general one. H-CLOSIB may eliminate redundant statistical information and the multi-scale version, M-CLOSIB, is more robust. We evaluated our method using four datasets: KTH TIPS (2-a) for material recognition, UIUC and USPTex for general texture recognition and JAFFE for face recognition. The results show that when we combine CLOSIB with well-known LBP-based descriptors, the hit rate increases in all the cases, introducing in this way the idea that CLOSIB can be used to enhance the description of texture in a significant number of situations. Additionally, a comparison with recent algorithms demonstrates that a combination of LBP methods with CLOSIB variants obtains comparable results to those of the state-of-the-art.

RESUMEN

BACKGROUND AND OBJECTIVE: Automatic classification of healthy tissues and organs based on histology images is an open problem, mainly due to the lack of automated tools. Solutions in this regard have potential in educational medicine and medical practices. Some preliminary advances have been made using image processing techniques and classical supervised learning. Due to the breakthrough performance of deep learning in various areas, we present an approach to recognise and classify, automatically, fundamental tissues and organs using Convolutional Neural Networks (CNN). METHODS: We adapt four popular CNNs architectures - ResNet, VGG19, VGG16 and Inception - to this problem through transfer learning. The resulting models are evaluated at three stages. Firstly, all the transferred networks are compared to each other. Secondly, the best resulting fine-tuned model is compared to an ad-hoc 2D multi-path model to outline the importance of transfer learning. Thirdly, the same model is evaluated against the state-of-the-art method, a cascade SVM using LBP-based descriptors, to contrast a traditional machine learning approach and a representation learning one. The evaluation task consists of separating six classes accurately: smooth muscle of the elastic artery, smooth muscle of the large vein, smooth muscle of the muscular artery, cardiac muscle, loose connective tissue, and light regions. The different networks are tuned on 6000 blocks of 100 × 100 pixels and tested on 7500. RESULTS: Our proposal yields F-score values between 0.717 and 0.928. The highest and lowest performances are for cardiac muscle and smooth muscle of the large vein, respectively. The main issue leading to limited classification scores for the latter class is its similarity with the elastic artery. However, this confusion is evidenced during manual annotation as well. Our algorithm reached improvements in F-score between 0.080 and 0.220 compared to the state-of-the-art machine learning approach. CONCLUSIONS: We conclude that it is possible to classify healthy cardiovascular tissues and organs automatically using CNNs and that deep learning holds great promise to improve tissue and organs classification. We left our training and test sets, models and source code publicly available to the research community.

Asunto(s)

RESUMEN

Textile based image retrieval for indoor environments can be used to retrieve images that contain the same textile, which may indicate that scenes are related. This makes up a useful approach for law enforcement agencies who want to find evidence based on matching between textiles. In this paper, we propose a novel pipeline that allows searching and retrieving textiles that appear in pictures of real scenes. Our approach is based on first obtaining regions containing textiles by using MSER on high pass filtered images of the RGB, HSV and Hue channels of the original photo. To describe the textile regions, we demonstrated that the combination of HOG and HCLOSIB is the best option for our proposal when using the correlation distance to match the query textile patch with the candidate regions. Furthermore, we introduce a new dataset, TextilTube, which comprises a total of 1913 textile regions labelled within 67 classes. We yielded 84.94% of success in the 40 nearest coincidences and 37.44% of precision taking into account just the first coincidence, which outperforms the current deep learning methods evaluated. Experimental results show that this pipeline can be used to set up an effective textile based image retrieval system in indoor environments.

RESUMEN

BACKGROUND: In this paper, we describe a histological ontology of the human cardiovascular system developed in collaboration among histology experts and computer scientists. RESULTS: The histological ontology is developed following an existing methodology using Conceptual Models (CMs) and validated using OOPS!, expert evaluation with CMs, and how accurately the ontology can answer the Competency Questions (CQ). It is publicly available at http://bioportal.bioontology.org/ontologies/HO and https://w3id.org/def/System . CONCLUSIONS: The histological ontology is developed to support complex tasks, such as supporting teaching activities, medical practices, and bio-medical research or having natural language interactions.

Asunto(s)

RESUMEN

BACKGROUND AND OBJECTIVE: Histological images have characteristics, such as texture, shape, colour and spatial structure, that permit the differentiation of each fundamental tissue and organ. Texture is one of the most discriminative features. The automatic classification of tissues and organs based on histology images is an open problem, due to the lack of automatic solutions when treating tissues without pathologies. METHOD: In this paper, we demonstrate that it is possible to automatically classify cardiovascular tissues using texture information and Support Vector Machines (SVM). Additionally, we realised that it is feasible to recognise several cardiovascular organs following the same process. The texture of histological images was described using Local Binary Patterns (LBP), LBP Rotation Invariant (LBPri), Haralick features and different concatenations between them, representing in this way its content. Using a SVM with linear kernel, we selected the more appropriate descriptor that, for this problem, was a concatenation of LBP and LBPri. Due to the small number of the images available, we could not follow an approach based on deep learning, but we selected the classifier who yielded the higher performance by comparing SVM with Random Forest and Linear Discriminant Analysis. Once SVM was selected as the classifier with a higher area under the curve that represents both higher recall and precision, we tuned it evaluating different kernels, finding that a linear SVM allowed us to accurately separate four classes of tissues: (i) cardiac muscle of the heart, (ii) smooth muscle of the muscular artery, (iii) loose connective tissue, and (iv) smooth muscle of the large vein and the elastic artery. The experimental validation was conducted using 3000 blocks of 100 × 100 sized pixels, with 600 blocks per class and the classification was assessed using a 10-fold cross-validation. RESULTS: using LBP as the descriptor, concatenated with LBPri and a SVM with linear kernel, the main four classes of tissues were recognised with an AUC higher than 0.98. A polynomial kernel was then used to separate the elastic artery and vein, yielding an AUC in both cases superior to 0.98. CONCLUSION: Following the proposed approach, it is possible to separate with very high precision (AUC greater than 0.98) the fundamental tissues of the cardiovascular system along with some organs, such as the heart, arteries and veins.

Asunto(s)

RESUMEN

Cardiovascular disease is the leading cause of death worldwide. Therefore, techniques for improving diagnosis and treatment in this field have become key areas for research. In particular, approaches for tissue image processing may support education system and medical practice. In this paper, an approach to automatic recognition and classification of fundamental tissues, using morphological information is presented. Taking a 40× or 10× histological image as input, three clusters are created with the k-means algorithm using a structural tensor and the red and the green channels. Loose connective tissue, light regions and cell nuclei are recognised on 40× images. Then, the cell nuclei's features - shape and spatial projection - and light regions are used to recognise and classify epithelial cells and tissue into flat, cubic and cylindrical. In a similar way, light regions, loose connective and muscle tissues are recognised on 10× images. Finally, the tissue's function and composition are used to refine muscle tissue recognition. Experimental validation is then carried out by histologist following expert criteria, along with manually annotated images that are used as a ground-truth. The results revealed that the proposed approach classified the fundamental tissues in a similar way to the conventional method employed by histologists. The proposed automatic recognition approach provides for epithelial tissues a sensitivity of 0.79 for cubic, 0.85 for cylindrical and 0.91 for flat. Furthermore, the experts gave our method an average score of 4.85 out of 5 in the recognition of loose connective tissue and 4.82 out of 5 for muscle tissue recognition.

Asunto(s)

RESUMEN

The assessment of the state of the acrosome is a priority in artificial insemination centres since it is one of the main causes of function loss. In this work, boar spermatozoa present in gray scale images acquired with a phase-contrast microscope have been classified as acrosome-intact or acrosome-damaged, after using fluorescent images for creating the ground truth. Based on shape prior criteria combined with Otsu's thresholding, regional minima and watershed transform, the spermatozoa heads were segmented and registered. One of the main novelties of this proposal is that, unlike what previous works stated, the obtained results show that the contour information of the spermatozoon head is important for improving description and classification. Other of this work novelties is that it confirms that combining different texture descriptors and contour descriptors yield the best classification rates for this problem up to date. The classification was performed with a Support Vector Machine backed by a Least Squares training algorithm and a linear kernel. Using the biggest acrosome intact-damaged dataset ever created, the early fusion approach followed provides a 0.9913 F-Score, outperforming all previous related works.

Asunto(s)

RESUMEN

The automated assessment of the sperm quality is an important challenge in the veterinary field. In this paper, we explore how to describe the acrosomes of boar spermatozoa using image analysis so that they can be automatically categorized as intact or damaged. Our proposal aims at characterizing the acrosomes by means of texture features. The texture is described using first order statistics and features derived from the co-occurrence matrix of the image, both computed from the original image and from the coefficients yielded by the Discrete Wavelet Transform. Texture descriptors are evaluated and compared with moments-based descriptors in terms of the classification accuracy they provide. Experimental results with a Multilayer Perceptron and the k-Nearest Neighbours classifiers show that texture descriptors outperform moment-based descriptors, reaching an accuracy of 94.93%, which makes this approach very attractive for the veterinarian community.

Asunto(s)

RESUMEN

We consider images of boar spermatozoa obtained with an optical phase-contrast microscope. Our goal is to automatically classify single sperm cells as acrosome-intact (class 1) or acrosome-damaged (class 2). Such classification is important for the estimation of the fertilization potential of a sperm sample for artificial insemination. We segment the sperm heads and compute a feature vector for each head. As a feature vector we use the gradient magnitude along the contour of the sperm head. We apply learning vector quantization (LVQ) to the feature vectors obtained for 320 heads that were labelled as intact or damaged using stains. A LVQ system with four prototypes (two for each class) allows us to classify cells with an overall test error of 6.8%. This is considered to be sufficient for semen quality control in an artificial insemination center.

Asunto(s)