Advancing Bag-of-visual-words Representations For Lesion Classification In Retinal Images.

Diabetic Retinopathy (DR) is a complication of diabetes that can lead to blindness if not readily discovered. Automated screening algorithms have the potential to improve identification of patients who need further medical attention. However, the identification of lesions must be accurate to be useful for clinical application. The bag-of-visual-words (BoVW) algorithm employs a maximum-margin classifier in a flexible framework that is able to detect the most common DR-related lesions such as microaneurysms, cotton-wool spots and hard exudates. BoVW allows to bypass the need for pre- and post-processing of the retinographic images, as well as the need of specific ad hoc techniques for identification of each type of lesion. An extensive evaluation of the BoVW model, using three large retinograph datasets (DR1, DR2 and Messidor) with different resolution and collected by different healthcare personnel, was performed. The results demonstrate that the BoVW classification approach can identify different lesions within an image without having to utilize different algorithms for each lesion reducing processing time and providing a more flexible diagnostic system. Our BoVW scheme is based on sparse low-level feature detection with a Speeded-Up Robust Features (SURF) local descriptor, and mid-level features based on semi-soft coding with max pooling. The best BoVW representation for retinal image classification was an area under the receiver operating characteristic curve (AUC-ROC) of 97.8% (exudates) and 93.5% (red lesions), applying a cross-dataset validation protocol. To assess the accuracy for detecting cases that require referral within one year, the sparse extraction technique associated with semi-soft coding and max pooling obtained an AUC of 94.2 ± 2.0%, outperforming current methods. Those results indicate that, for retinal image classification tasks in clinical practice, BoVW is equal and, in some instances, surpasses results obtained using dense detection (widely believed to be the best choice in many vision problems) for the low-level descriptors.

Automatic top-down processing explains common left occipito-temporal responses to visual words and objects.

Previous studies have demonstrated that a region in the left ventral occipito-temporal (LvOT) cortex is highly selective to the visual forms of written words and objects relative to closely matched visual stimuli. Here, we investigated why LvOT activation is not higher for reading than picture naming even though written words and pictures of objects have grossly different visual forms. To compare neuronal responses for words and pictures within the same LvOT area, we used functional magnetic resonance imaging adaptation and instructed participants to name target stimuli that followed briefly presented masked primes that were either presented in the same stimulus type as the target (word-word, picture-picture) or a different stimulus type (picture-word, word-picture). We found that activation throughout posterior and anterior parts of LvOT was reduced when the prime had the same name/response as the target irrespective of whether the prime-target relationship was within or between stimulus type. As posterior LvOT is a visual form processing area, and there was no visual form similarity between different stimulus types, we suggest that our results indicate automatic top-down influences from pictures to words and words to pictures. This novel perspective motivates further investigation of the functional properties of this intriguing region.

Electrostimulation mapping of comprehension of auditory and visual words.

In order to spare functional areas during the removal of brain tumours, electrical stimulation mapping was used in 90 patients (77 in the left hemisphere and 13 in the right; 2754 cortical sites tested). Language functions were studied with a special focus on comprehension of auditory and visual words and the semantic system. In addition to naming, patients were asked to perform pointing tasks from auditory and visual stimuli (using sets of 4 different images controlled for familiarity), and also auditory object (sound recognition) and Token test tasks. Ninety-two auditory comprehension interference sites were observed. We found that the process of auditory comprehension involved a few, fine-grained, sub-centimetre cortical territories. Early stages of speech comprehension seem to relate to two posterior regions in the left superior temporal gyrus. Downstream lexical-semantic speech processing and sound analysis involved 2 pathways, along the anterior part of the left superior temporal gyrus, and posteriorly around the supramarginal and middle temporal gyri. Electrostimulation experimentally dissociated perceptual consciousness attached to speech comprehension. The initial word discrimination process can be considered as an "automatic" stage, the attention feedback not being impaired by stimulation as would be the case at the lexical-semantic stage. Multimodal organization of the superior temporal gyrus was also detected since some neurones could be involved in comprehension of visual material and naming. These findings demonstrate a fine graded, sub-centimetre, cortical representation of speech comprehension processing mainly in the left superior temporal gyrus and are in line with those described in dual stream models of language comprehension processing.

Image categorization through optimum path forest and visual words

Different from the first attempts to solve the image categorization problem (often based on global features), recently, several researchers have been tackling this research branch through a new vantage point - using features around locally invariant interest points and visual dictionaries. Although several advances have been done in the visual dictionaries literature in the past few years, a problem we still need to cope with is calculation of the number of representative words in the dictionary. Therefore, in this paper we introduce a new solution for automatically finding the number of visual words in an N-Way image categorization problem by means of supervised pattern classification based on optimum-path forest. © 2011 IEEE.

Classificazione di oggetti in immagini attraverso il modello Bag of Visual Words

Generic object recognition is an important function of the human visual system and everybody finds it highly useful in their everyday life. For an artificial vision system it is a really hard, complex and challenging task because instances of the same object category can generate very different images, depending of different variables such as illumination conditions, the pose of an object, the viewpoint of the camera, partial occlusions, and unrelated background clutter. The purpose of this thesis is to develop a system that is able to classify objects in 2D images based on the context, and identify to which category the object belongs to. Given an image, the system can classify it and decide the correct categorie of the object. Furthermore the objective of this thesis is also to test the performance and the precision of different supervised Machine Learning algorithms in this specific task of object image categorization. Through different experiments the implemented application reveals good categorization performances despite the difficulty of the problem. However this project is open to future improvement; it is possible to implement new algorithms that has not been invented yet or using other techniques to extract features to make the system more reliable. This application can be installed inside an embedded system and after trained (performed outside the system), so it can become able to classify objects in a real-time. The information given from a 3D stereocamera, developed inside the department of Computer Engineering of the University of Bologna, can be used to improve the accuracy of the classification task. The idea is to segment a single object in a scene using the depth given from a stereocamera and in this way make the classification more accurate.

Integrated visual vocabulary in latent Dirichlet allocation-based scene classification for IKONOS image

Scene classification based on latent Dirichlet allocation (LDA) is a more general modeling method known as a bag of visual words, in which the construction of a visual vocabulary is a crucial quantization process to ensure success of the classification. A framework is developed using the following new aspects: Gaussian mixture clustering for the quantization process, the use of an integrated visual vocabulary (IVV), which is built as the union of all centroids obtained from the separate quantization process of each class, and the usage of some features, including edge orientation histogram, CIELab color moments, and gray-level co-occurrence matrix (GLCM). The experiments are conducted on IKONOS images with six semantic classes (tree, grassland, residential, commercial/industrial, road, and water). The results show that the use of an IVV increases the overall accuracy (OA) by 11 to 12% and 6% when it is implemented on the selected and all features, respectively. The selected features of CIELab color moments and GLCM provide a better OA than the implementation over CIELab color moment or GLCM as individuals. The latter increases the OA by only ∼2 to 3%. Moreover, the results show that the OA of LDA outperforms the OA of C4.5 and naive Bayes tree by ∼20%. © 2014 Society of Photo-Optical Instrumentation Engineers (SPIE) [DOI: 10.1117/1.JRS.8.083690]

Automatic visual dictionary generation through Optimum-Path Forest clustering

Image categorization by means of bag of visual words has received increasing attention by the image processing and vision communities in the last years. In these approaches, each image is represented by invariant points of interest which are mapped to a Hilbert Space representing a visual dictionary which aims at comprising the most discriminative features in a set of images. Notwithstanding, the main problem of such approaches is to find a compact and representative dictionary. Finding such representative dictionary automatically with no user intervention is an even more difficult task. In this paper, we propose a method to automatically find such dictionary by employing a recent developed graph-based clustering algorithm called Optimum-Path Forest, which does not make any assumption about the visual dictionary's size and is more efficient and effective than the state-of-the-art techniques used for dictionary generation. © 2012 IEEE.

Automatic visual dictionary generation through optimum-path forest clusteringautomatic visual dictionary generation through optimum-path forest clustering

Image categorization by means of bag of visual words has received increasing attention by the image processing and vision communities in the last years. In these approaches, each image is represented by invariant points of interest which are mapped to a Hilbert Space representing a visual dictionary which aims at comprising the most discriminative features in a set of images. Notwithstanding, the main problem of such approaches is to find a compact and representative dictionary. Finding such representative dictionary automatically with no user intervention is an even more difficult task. In this paper, we propose a method to automatically find such dictionary by employing a recent developed graph-based clustering algorithm called Optimum-Path Forest, which does not make any assumption about the visual dictionary's size and is more efficient and effective than the state-of-the-art techniques used for dictionary generation.

Invariance of brain-wave representations of simple visual images and their names

In two experiments, electric brain waves of 14 subjects were recorded under several different conditions to study the invariance of brain-wave representations of simple patches of colors and simple visual shapes and their names, the words blue, circle, etc. As in our earlier work, the analysis consisted of averaging over trials to create prototypes and test samples, to both of which Fourier transforms were applied, followed by filtering and an inverse transformation to the time domain. A least-squares criterion of fit between prototypes and test samples was used for classification. The most significant results were these. By averaging over different subjects, as well as trials, we created prototypes from brain waves evoked by simple visual images and test samples from brain waves evoked by auditory or visual words naming the visual images. We correctly recognized from 60% to 75% of the test-sample brain waves. The general conclusion is that simple shapes such as circles and single-color displays generate brain waves surprisingly similar to those generated by their verbal names. These results, taken together with extensive psychological studies of auditory and visual memory, strongly support the solution proposed for visual shapes, by Bishop Berkeley and David Hume in the 18th century, to the long-standing problem of how the mind represents simple abstract ideas.

Scene Classification Using a Hybrid Generative/Discriminative Approach

We investigate whether dimensionality reduction using a latent generative model is beneficial for the task of weakly supervised scene classification. In detail, we are given a set of labeled images of scenes (for example, coast, forest, city, river, etc.), and our objective is to classify a new image into one of these categories. Our approach consists of first discovering latent ";topics"; using probabilistic Latent Semantic Analysis (pLSA), a generative model from the statistical text literature here applied to a bag of visual words representation for each image, and subsequently, training a multiway classifier on the topic distribution vector for each image. We compare this approach to that of representing each image by a bag of visual words vector directly and training a multiway classifier on these vectors. To this end, we introduce a novel vocabulary using dense color SIFT descriptors and then investigate the classification performance under changes in the size of the visual vocabulary, the number of latent topics learned, and the type of discriminative classifier used (k-nearest neighbor or SVM). We achieve superior classification performance to recent publications that have used a bag of visual word representation, in all cases, using the authors' own data sets and testing protocols. We also investigate the gain in adding spatial information. We show applications to image retrieval with relevance feedback and to scene classification in videos

Scene Classification Using a Hybrid Generative/Discriminative Approach

We investigate whether dimensionality reduction using a latent generative model is beneficial for the task of weakly supervised scene classification. In detail, we are given a set of labeled images of scenes (for example, coast, forest, city, river, etc.), and our objective is to classify a new image into one of these categories. Our approach consists of first discovering latent ";topics"; using probabilistic Latent Semantic Analysis (pLSA), a generative model from the statistical text literature here applied to a bag of visual words representation for each image, and subsequently, training a multiway classifier on the topic distribution vector for each image. We compare this approach to that of representing each image by a bag of visual words vector directly and training a multiway classifier on these vectors. To this end, we introduce a novel vocabulary using dense color SIFT descriptors and then investigate the classification performance under changes in the size of the visual vocabulary, the number of latent topics learned, and the type of discriminative classifier used (k-nearest neighbor or SVM). We achieve superior classification performance to recent publications that have used a bag of visual word representation, in all cases, using the authors' own data sets and testing protocols. We also investigate the gain in adding spatial information. We show applications to image retrieval with relevance feedback and to scene classification in videos

HOW FAR do WE GET USING MACHINE LEARNING BLACK-BOXES?

Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP)

Composição de dicionários visuais utilizando agrupamento de dados por Florestas de Caminhos Ótimos

Pós-graduação em Ciência da Computação - IBILCE

Segmentazione e categorizzazione di oggetti mediante immagini depth e deep learning

Questo lavoro è iniziato con uno studio teorico delle principali tecniche di classificazione di immagini note in letteratura, con particolare attenzione ai più diffusi modelli di rappresentazione dell’immagine, quali il modello Bag of Visual Words, e ai principali strumenti di Apprendimento Automatico (Machine Learning). In seguito si è focalizzata l’attenzione sulla analisi di ciò che costituisce lo stato dell’arte per la classificazione delle immagini, ovvero il Deep Learning. Per sperimentare i vantaggi dell’insieme di metodologie di Image Classification, si è fatto uso di Torch7, un framework di calcolo numerico, utilizzabile mediante il linguaggio di scripting Lua, open source, con ampio supporto alle metodologie allo stato dell’arte di Deep Learning. Tramite Torch7 è stata implementata la vera e propria classificazione di immagini poiché questo framework, grazie anche al lavoro di analisi portato avanti da alcuni miei colleghi in precedenza, è risultato essere molto efficace nel categorizzare oggetti in immagini. Le immagini su cui si sono basati i test sperimentali, appartengono a un dataset creato ad hoc per il sistema di visione 3D con la finalità di sperimentare il sistema per individui ipovedenti e non vedenti; in esso sono presenti alcuni tra i principali ostacoli che un ipovedente può incontrare nella propria quotidianità. In particolare il dataset si compone di potenziali ostacoli relativi a una ipotetica situazione di utilizzo all’aperto. Dopo avere stabilito dunque che Torch7 fosse il supporto da usare per la classificazione, l’attenzione si è concentrata sulla possibilità di sfruttare la Visione Stereo per aumentare l’accuratezza della classificazione stessa. Infatti, le immagini appartenenti al dataset sopra citato sono state acquisite mediante una Stereo Camera con elaborazione su FPGA sviluppata dal gruppo di ricerca presso il quale è stato svolto questo lavoro. Ciò ha permesso di utilizzare informazioni di tipo 3D, quali il livello di depth (profondità) di ogni oggetto appartenente all’immagine, per segmentare, attraverso un algoritmo realizzato in C++, gli oggetti di interesse, escludendo il resto della scena. L’ultima fase del lavoro è stata quella di testare Torch7 sul dataset di immagini, preventivamente segmentate attraverso l’algoritmo di segmentazione appena delineato, al fine di eseguire il riconoscimento della tipologia di ostacolo individuato dal sistema.

A Food Recognition System for Diabetic Patients based on an Optimized Bag of Features Model

Computer vision-based food recognition could be used to estimate a meal's carbohydrate content for diabetic patients. This study proposes a methodology for automatic food recognition, based on the Bag of Features (BoF) model. An extensive technical investigation was conducted for the identification and optimization of the best performing components involved in the BoF architecture, as well as the estimation of the corresponding parameters. For the design and evaluation of the prototype system, a visual dataset with nearly 5,000 food images was created and organized into 11 classes. The optimized system computes dense local features, using the scale-invariant feature transform on the HSV color space, builds a visual dictionary of 10,000 visual words by using the hierarchical k-means clustering and finally classifies the food images with a linear support vector machine classifier. The system achieved classification accuracy of the order of 78%, thus proving the feasibility of the proposed approach in a very challenging image dataset.