103 research outputs found
Breast Lesion Detection from Mammograms Using Deep Convolutional Neural Networks
No full textMammography has a central role in screening and diagnosis of breast lesions, allowing early detection of the pathology and reduction of fatal cases. Deep Convolutional Neural Networks have shown a great potentiality to address the issue of early detection of breast cancer with an acceptable level of accuracy and reproducibility. In the present paper, we illustrate the development of a deep learning study aimed to process and classify lesions in mammograms with the use of slender neural networks not yet used in literature. For this reason, a traditional convolution network was compared with a novel one obtained making use of much more efficient depth wise separable convolution layers. Preliminary numerical results are detailed and future plans outlined
A framework for interpreting, modeling and recognizing human body gestures through 3D eigenpostures
No full textBiometric Signals Estimation Using Single Photon Camera and Deep Learning
No full textThe problem of performing remote biomedical measurements using just a video stream of a subject face is called remote photoplethysmography (rPPG). The aim of this work is to propose a novel method able to perform rPPG using single-photon avalanche diode (SPAD) cameras. These are extremely accurate cameras able to detect even a single photon and are already used in many other applications. Moreover, a novel method that mixes deep learning and traditional signal analysis is proposed in order to extract and study the pulse signal. Experimental results show that this system achieves accurate results in the estimation of biomedical information such as heart rate, respiration rate, and tachogram. Lastly, thanks to the adoption of the deep learning segmentation method and dependability checks, this method could be adopted in non-ideal working conditions—for example, in the presence of partial facial occlusions
Accurate characterization of embedded Structure from Motion
No full textTrajectory estimation and 3d scene reconstruction from single camera, e.g. Structure from Motion, is going to have a central role in the future of automotive industry. Typical appliance fields will be: Collisions avoidance with any kind of object (people included), parking assisted maneuvers and many more. Indeed various countries are becoming more and more concerned about road traffic safety and therefore through its 'Advanced Program', EuroNCAP rewards vehicle manufacturers who employ Advanced Safety Technologies that assists the driver. This paper had mainly two different goals: (1) to describe the implementation of a state of art Structure from Motion pipeline able to run in real time with embedded fish-eye camera, which includes nonlinear optimization (i.e. local bundle adjustment); (2) to demonstrate quantitatively its performances on a synthetic test space specifically designed for its characterization in term of accuracy
Deep Skin Detection on Low Resolution Grayscale Images
No full textIn this work we present a facial skin detection method, based on a deep learning architecture, that is able to precisely associate a skin label to each pixel of a given image depicting a face. This is an important preliminary step in many applications, such as remote photoplethysmography (rPPG) in which the hearth rate of a subject needs to be estimated analyzing a video of his/her face. The proposed method can detect skin pixels even in low resolution grayscale face images (64 × 32 pixel). A dataset is also described and proposed in order to train the deep learning model. Given the small amount of data available, a transfer learning approach is adopted and validated in order to learn to solve the skin detection problem exploiting a colorization network. Qualitative and quantitative results are reported testing the method on different datasets and in presence of general illumination, facial expressions, object occlusions and it is able to work regardless of the gender, age and ethnicity of the subject
Visual Odometry from Omnidirectional Images for Intelligent Transportation
No full textIn this article we use omnidirectional images obtained from equirectangular panoramas of Google MapsTM to estimate camera egomotion. The systems was also tested using a 360 camera. The goal is to provide an effective and accurate positioning system for indoor environments or in urban canyons where GPS signal could be absent. We reformulated classical Computer Vision geometrical constraints for pin-hole cameras, like epipolar and trifocal tensor, to omnidirectional cameras obtaining new and effective equations to accurately reconstruct the camera path using couples or triplets of omnidirectional images. Tests have been performed on straight and curved paths to validate the presented approaches
Visual Search of multiple objects from a single query
No full textHundreds of millions of images are uploaded to the cloud every day. Innovative applications able to analyze and extract efficiently information from such a big database are needed nowadays more than ever. Visual search is an application able to retrieve information of a query image comparing it against a large image database. In this paper a Visual Search pipeline implementation is presented able to retrieve multiple objects depicted in a single query image. Quantitative and qualitative precision results are shown on both real and synthetic datasets
Dissection of genetic associations with language-related traits in population-based cohorts
No full textThe author was supported by the Wellcome Trust [076566/Z/05/Z]; [075491/Z/04] and the Medical Research Council [G0800523/86473].Recent advances in the field of language-related disorders have led to the identification of candidate genes for specific language impairment (SLI) and dyslexia. Replication studies have been conducted in independent samples including population-based cohorts, which can be characterised for a large number of relevant cognitive measures. The availability of a wide range of phenotypes allows us to not only identify the most suitable traits for replication of genetic association but also to refine the associated cognitive trait. In addition, it is possible to test for pleiotropic effects across multiple phenotypes which could explain the extensive comorbidity observed across SLI, dyslexia and other neurodevelopmental disorders. The availability of genome-wide genotype data for such cohorts will facilitate this kind of analysis but important issues, such as multiple test corrections, have to be taken into account considering that small effect sizes are expected to underlie such associations.Peer reviewe
Hyperspectral X-ray denoising: Model-based and data-driven solutions
No full textIn this paper we deal with the problem of hyperspectral X-Ray image denoising. In particular, we compare a classical model-based Wiener filter solution with a data-driven methodology based on a Convolutional Autoencoder. A challenging aspect is related to the specific kind of 2D signal we are processing: it presents mixed dimensions information since on the vertical axis there is the pixels position while, on the abscissa, there are the different wavelengths associated to the acquired X-Ray spectrum. The goal is to approximate the denoising function using a learning-from-data approach and to verify its capability to emulate the Wiener filter using a much less demanding approach in terms of signal and noise statistical knowledge. We show that, after training, the CNN is able to properly restore the 2D signal with results very close to the Wiener filter, honouring the proper signal shape
- …
