Deep generative models for facial keypoints detection

A new area of machine learning research called deep learning, has moved machine learning closer to one of its original goals: artificial intelligence and general learning algorithm. The key idea is to pretrain models in completely unsupervised way and finally they can be fine-tuned for the task at hand using supervised learning. In this thesis, a general introduction to deep learning models and algorithms are given and these methods are applied to facial keypoints detection. The task is to predict the positions of 15 keypoints on grayscale face images. Each predicted keypoint is specified by an (x,y) real-valued pair in the space of pixel indices. In experiments, we pretrained deep belief networks (DBN) and finally performed a discriminative fine-tuning. We varied the depth and size of an architecture. We tested both deterministic and sampled hidden activations and the effect of additional unlabeled data on pretraining. The experimental results show that our model provides better results than publicly available benchmarks for the dataset.

Feature Point Detection and Curve Approximation for Early Processing of Freehand Sketches

Freehand sketching is both a natural and crucial part of design, yet is unsupported by current design automation software. We are working to combine the flexibility and ease of use of paper and pencil with the processing power of a computer to produce a design environment that feels as natural as paper, yet is considerably smarter. One of the most basic steps in accomplishing this is converting the original digitized pen strokes in the sketch into the intended geometric objects using feature point detection and approximation. We demonstrate how multiple sources of information can be combined for feature detection in strokes and apply this technique using two approaches to signal processing, one using simple average based thresholding and a second using scale space.

Feature-point detection using distance transforms: Application to tracking tropical convective complexes

The identification, tracking, and statistical analysis of tropical convective complexes using satellite imagery is explored in the context of identifying feature points suitable for tracking. The feature points are determined based on the shape of complexes using the distance transform technique. This approach has been applied to the determination feature points for tropical convective complexes identified in a time series of global cloud imagery. The feature points are used to track the complexes, and from the tracks statistical diagnostic fields are computed. This approach allows the nature and distribution of organized deep convection in the Tropics to be explored.

Change-point detection in environmental time series based on the informational approach

In this study, the Schwarz Information Criterion (SIC) is applied in order to detect change-points in the time series of surface water quality variables. The application of change-point analysis allowed detecting change-points in both the mean and the variance in series under study. Time variations in environmental data are complex and they can hinder the identification of the so-called change-points when traditional models are applied to this type of problems. The assumptions of normality and uncorrelation are not present in some time series, and so, a simulation study is carried out in order to evaluate the methodology’s performance when applied to non-normal data and/or with time correlation.

Detecció automàtica de punts facials característics mitjançant un algoritme inspirat en el còrtex cerebral

En aquest article presentem l'aplicació d'un algoritme inspirat en el còrtex cerebral anomenat HMAX com a part central d'un procés de detecció automàtica de punts facials característics en imatges.

A comparison of face and facial feature detectors based on the viola-jones general object detection framework

[EN]The human face provides useful information during interaction; therefore, any system integrating Vision- BasedHuman Computer Interaction requires fast and reliable face and facial feature detection. Different approaches have focused on this ability but only open source implementations have been extensively used by researchers. A good example is the Viola–Jones object detection framework that particularly in the context of facial processing has been frequently used.

Microbiological diagnosis of community-acquired respiratory tract infections by nucleic acid detection.

Background: Microbiological diagnostic procedures have changed significantly over the last decade. Initially the implementation of the polymerase chain reaction (PCR) resulted in improved detection tests for microbes that were difficult or even impossible to detect by conventional methods such as culture and serology, especially in community-acquired respiratory tract infections (CA-RTI). A further improvement was the development of real-time PCR, which allows end point detection and quantification, and many diagnostic laboratories have now implemented this powerful method. Objective: At present, new performant and convenient molecular tests have emerged targeting in parallel many viruses and bacteria responsible for lower and/or upper respiratory tract infections. The range of test formats and microbial agents detected is evolving very quickly and the added value of these new tests needs to be studied in terms of better use of antibiotics, better patient management, duration of hospitalization and overall costs. Conclusions: Molecular tools for a better microbial documentation of CA-RTI are now available. Controlled studies are now required to address the relevance issue of these new methods, such as, for example, the role of some newly detected respiratory viruses or of the microbial DNA load in a particular patient at a particular time. The future challenge for molecular diagnosis will be to become easy to handle, highly efficient and cost-effective, delivering rapid results with a direct impact on clinical management.

Supervised localimage feature detection

This thesis is about detection of local image features. The research topic belongs to the wider area of object detection, which is a machine vision and pattern recognition problem where an object must be detected (located) in an image. State-of-the-art object detection methods often divide the problem into separate interest point detection and local image description steps, but in this thesis a different technique is used, leading to higher quality image features which enable more precise localization. Instead of using interest point detection the landmark positions are marked manually. Therefore, the quality of the image features is not limited by the interest point detection phase and the learning of image features is simplified. The approach combines both interest point detection and local description into one phase for detection. Computational efficiency of the descriptor is therefore important, leaving out many of the commonly used descriptors as unsuitably heavy. Multiresolution Gabor features has been the main descriptor in this thesis and improving their efficiency is a significant part. Actual image features are formed from descriptors by using a classifierwhich can then recognize similar looking patches in new images. The main classifier is based on Gaussian mixture models. Classifiers are used in one-class classifier configuration where there are only positive training samples without explicit background class. The local image feature detection method has been tested with two freely available face detection databases and a proprietary license plate database. The localization performance was very good in these experiments. Other applications applying the same under-lying techniques are also presented, including object categorization and fault detection.

Change points detection in crime-related time series: an on-line fuzzy approach based on a shape space representation

The extension of traditional data mining methods to time series has been effectively applied to a wide range of domains such as finance, econometrics, biology, security, and medicine. Many existing mining methods deal with the task of change points detection, but very few provide a flexible approach. Querying specific change points with linguistic variables is particularly useful in crime analysis, where intuitive, understandable, and appropriate detection of changes can significantly improve the allocation of resources for timely and concise operations. In this paper, we propose an on-line method for detecting and querying change points in crime-related time series with the use of a meaningful representation and a fuzzy inference system. Change points detection is based on a shape space representation, and linguistic terms describing geometric properties of the change points are used to express queries, offering the advantage of intuitiveness and flexibility. An empirical evaluation is first conducted on a crime data set to confirm the validity of the proposed method and then on a financial data set to test its general applicability. A comparison to a similar change-point detection algorithm and a sensitivity analysis are also conducted. Results show that the method is able to accurately detect change points at very low computational costs. More broadly, the detection of specific change points within time series of virtually any domain is made more intuitive and more understandable, even for experts not related to data mining.

Bayesian change-point analysis in linear regression model with scale mixtures of normal distributions

In this thesis, we consider Bayesian inference on the detection of variance change-point models with scale mixtures of normal (for short SMN) distributions. This class of distributions is symmetric and thick-tailed and includes as special cases: Gaussian, Student-t, contaminated normal, and slash distributions. The proposed models provide greater flexibility to analyze a lot of practical data, which often show heavy-tail and may not satisfy the normal assumption. As to the Bayesian analysis, we specify some prior distributions for the unknown parameters in the variance change-point models with the SMN distributions. Due to the complexity of the joint posterior distribution, we propose an efficient Gibbs-type with Metropolis- Hastings sampling algorithm for posterior Bayesian inference. Thereafter, following the idea of [1], we consider the problems of the single and multiple change-point detections. The performance of the proposed procedures is illustrated and analyzed by simulation studies. A real application to the closing price data of U.S. stock market has been analyzed for illustrative purposes.

Método potenciométrico para determinação de cobre em cachaça

'Cachaça' is the Brazilian name for the spirit obtained from sugarcane. According to Brazilian regulations, it may be sold raw or with addition of sugar and may contain up to 5 mg/L of copper. Copper in "cachaça" was determined by titration with EDTA, using a homemade copper membrane electrode for end-point detection. It was found a pooled standard deviation of 0,057 mg/L and there was no significant difference between the results obtained by the potentiometric method and by flame atomic absorption spectrometry with standard addition. Among the 21 'cachaça' samples from 16 different brands analyzed, three overpassed the legal copper limit. For its characteristics of accuracy, precision, and speed, the potentiometric method may be employed advantageously in routine analysis, specially when low cost is a major concern.

Método titrimétrico para determinar fosfito em amostras agroindustriais

A new titrimetric method for the determination of phosphite in fertilizer samples, based on reaction of H3PO3 with standard iodine solution in neutral media, is proposed. Diluted samples containing ca. 0.4% m/v P2O5 are heated and titrated with 0.05 mol L-1 iodine standard until the solution becomes faint yellow. Back titration is also feasible: a slight excess of titrant is added followed by starch indicator and titration is completed taking as the end point the change in color from blue to colorless. The influence of chemical composition and pH of buffers, temperature and foreign species on waiting time and end-point detection were investigated. For the Na2HPO4/NaH2PO4 buffer (pH 6.8) at 70 °C, the titration time was 10 min, corresponding to about 127 mg iodine, 200 mg KI and 174 mg Na2HPO4 and 176 mg NaH2PO4 consumed per determination. Accuracy was checked for phosphite determination in seven fertilizer samples. Results obtained by the proposed procedure were in agreement with those obtained by spectrophotometry at 95% confidence level. The R.S.D. (n=10) for direct and back titration was 0.4% and 1.3% respectively.

A Contour-based Separation of VerticallyAttached Traffic Signs

This report presents an algorithm for locating the cut points for and separatingvertically attached traffic signs in Sweden. This algorithm provides severaladvanced digital image processing features: binary image which representsvisual object and its complex rectangle background with number one and zerorespectively, improved cross correlation which shows the similarity of 2Dobjects and filters traffic sign candidates, simplified shape decompositionwhich smoothes contour of visual object iteratively in order to reduce whitenoises, flipping point detection which locates black noises candidates, chasmfilling algorithm which eliminates black noises, determines the final cut pointsand separates originally attached traffic signs into individual ones. At each step,the mediate results as well as the efficiency in practice would be presented toshow the advantages and disadvantages of the developed algorithm. Thisreport concentrates on contour-based recognition of Swedish traffic signs. Thegeneral shapes cover upward triangle, downward triangle, circle, rectangle andoctagon. At last, a demonstration program would be presented to show howthe algorithm works in real-time environment.

Método titrimétrico para determinar fosfito em amostras agroindustriais

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