Le développement de méthodes de la détection en temps réel de cellules anormales (pouvant être considérées comme des cellules cancéreuses) par captures et traitements bio-images sont des axes de recherche importants dans le domaine biomédical car cela contribue à diagnostiquer un cancer. C’est dans ce contexte que se situe ces travaux de thèse. Plus précisément, les travaux présentés dans ce manuscrit, se focalise sur le développement de procédures de lecture, de détection et de classification automatiques de bio-images de cellules anormales considérées comme des cellules cancéreuses. Par conséquent, une première étape du travail à consister à déterminer une solution de détection, à partir d’images microscopiques multispectrales permettant une répétitivité d’images sur une gamme de longueurs d'ondes de certains types de bio-images anormales associées à différents stades ou évolutions de cellules cancéreuses. L’approche développée dans ces travaux repose sur l’exploitation d’une nouvelle méthode de segmentation basée sur l'intensité de la couleur et pouvant être appliquée sur des séquences d'objets dans une image en reformant de manière adaptative et itérative la localisation et la couverture de contours réels de cellules. Cette étape préalable de segmentation est primordiale et permet une classification des tissus anormaux en utilisant la méthode de réseau de neurones à convolution (CNN) appliqué sur les images microscopiques segmenté de type snake. L’approche permet d’obtenir de bas résultats comparativement à une approche basée sur d’autres méthodes de segmentation de la littérature. En effet, cette méthode de classification atteint des valeurs de performance de 100% pour la phase d’apprentissage et de 99.168 % pour les phases de test. Cette méthode est comparée à différents travaux antérieurs et basée sur différentes fonctionnalités d'extraction, et a prouvé son efficacité par rapport à ces autres méthodes. En terme de perspectives, les travaux futurs visent à valider notre approche sur des ensembles de données plus larges, et à explorer différentes architectures CNN selon différents critères d’optimisation / Development of methods for help diagnosis of the real time detection of abnormal cells (which can be considered as cancer cells) through bio-image processing and detection are most important research directions in information science and technology. Our work has been concerned by developing automatic reading procedures of the normal and abnormal bio-images tissues. Therefore, the first step of our work is to detect a certain type of abnormal bio-images associated to many types evolution of cancer within a Microscopic multispectral image, which is an image, repeated in many wavelengths. And using a new segmentation method that reforms itself in an iterative adaptive way to localize and cover the real cell contour, using some segmentation techniques. It is based on color intensity and can be applied on sequences of objects in the image. This work presents a classification of the abnormal tissues using the Convolution neural network (CNN), where it was applied on the microscopic images segmented using the snake method, which gives a high performance result with respect to the other segmentation methods. This classification method reaches high performance values, where it reaches 100% for training and 99.168% for testing. This method was compared to different papers that uses different feature extraction, and proved its high performance with respect to other methods. As a future work, we will aim to validate our approach on a larger datasets, and to explore different CNN architectures and the optimization of the hyper-parameters, in order to increase its performance, and it will be applied to relevant medical imaging tasks including computer-aided diagnosis

<p>In recent years, deep learning models have become popular in the real-time embedded application, but there are many complexities for hardware deployment because of limited resources such as memory, computational power, and energy. Recent research in the field of deep learning focuses on reducing the model size of the Convolution Neural Network (CNN) by various compression techniques like Architectural compression, Pruning, Quantization, and Encoding (e.g., Huffman encoding). Network pruning is one of the promising technique to solve these problems.</p> <p>This thesis proposes methods to prune the convolution neural network (SqueezeNet) without introducing network sparsity in the pruned model. </p> <p>This thesis proposes three methods to prune the CNN to decrease the model size of CNN without a significant drop in the accuracy of the model.</p> <p>1: Pruning based on Taylor expansion of change in cost function Delta C.</p> <p>2: Pruning based on L₂ normalization of activation maps.</p> <p>3: Pruning based on a combination of method 1 and method 2.</p><p>The proposed methods use various ranking methods to rank the convolution kernels and prune the lower ranked filters afterwards SqueezeNet model is fine-tuned by backpropagation. Transfer learning technique is used to train the SqueezeNet on the CIFAR-10 dataset. Results show that the proposed approach reduces the SqueezeNet model by 72% without a significant drop in the accuracy of the model (optimal pruning efficiency result). Results also show that Pruning based on a combination of Taylor expansion of the cost function and L₂ normalization of activation maps achieves better pruning efficiency compared to other individual pruning criteria and most of the pruned kernels are from mid and high-level layers. The Pruned model is deployed on BlueBox 2.0 using RTMaps software and model performance was evaluated.</p><p></p>

The two main bottlenecks using deep neural networks are data dependency and training time. This thesis proposes a novel method for weight initialization of the convolutional layers in a convolutional neural network. This thesis introduces the usage of sparse dictionaries. A sparse dictionary optimized on domain specific data can be seen as a set of intelligent feature extracting filters. This thesis investigates the effect of using such filters as kernels in the convolutional layers in the neural network. How do they affect the training time and final performance? The dataset used here is the Cityscapes-dataset which is a library of 25000 labeled road scene images.The sparse dictionary was acquired using the K-SVD method. The filters were added to two different networks whose performance was tested individually. One of the architectures is much deeper than the other. The results have been presented for both networks. The results show that filter initialization is an important aspect which should be taken into consideration while training the deep networks for semantic segmentation.

In the scheme of Non Destructive Testing (NDT), defect detection is an important process. Traditional image processing techniques have successfully been used for defect recognition. Usage of machine learning techniques is still in the initial stages of development. Convolution Neural Networks (CNN) is widely used for object classification one such scenario is defect classification in weld tubes. With the advent of deep learning techniques such as transfer learning, we can transfer knowledge gained in one domain successfully into other. Pre-trained models successfully learn features from large scale datasets that can be used for in domains having sparse data and smaller datasets. The aim of this work is to help a manual inspector in recognition of defects on the weld tubes. With a given set of images, we proceed by forming unique pipeline architecture for automatic defect recognition. The research in this thesis focuses on extraction of welds using image segmentation techniques, creating a dataset of defects and using it to on pre-trained Convolution Neural Networks of VGG16, VGG19, Inception V3 and ResNet101. We evaluate the models on different metrics finding the best suited model for the created dataset. Further a prototype sliding window solution is used to find defects over the extracted weld region. We also present the limitations of this approach and suggest modifications that can be implemented in the future.

This work focuses on a theoretical explanation of heart rhythm disorders and the possibility of their automatic detection using deep learning networks. For the purposes of this work, a total of 6884 10-second ECG recordings with measured eight leads were used. Those recordings were divided into 5 groups according to heart rhythm into a group of records with atrial fibrillation, sinus rhythms, supraventricular rhythms, ventricular rhythms, and the last group consisted of the others records. Individual groups were unbalanced represented and more than 85 % of the total number of data are sinus rhythm group records. The used classification methods served effectively as a record detector of the largest group and the most effective of all was a procedure consisting of a 2D convolutional neural network into which data entered in the form of scalalograms (classification procedure number 3). It achieved results of precision of 91%, recall of 96% and F1-score values of 0.93. On the contrary, when classifying all groups at the same time, there were no such quality results for all groups. The most efficient procedure seems to be a variant composed of PCA on eight input signals with the gain of one output signal, which becomes the input of a 1D convolutional neural network (classification procedure number 5). This procedure achieved the following F1-score values: 1) group of records with atrial fibrillation 0.54, 2) group of sinus rhythms 0.91, 3) group of supraventricular rhythms 0.65, 4) group of ventricular rhythms 0.68, 5) others records 0.65.

This master’s thesis deals with designs and implementation of systems for music cover recognition. The introduction part is devoted to the calculation parameters from audio signal using Music Information Retrieval techniques. Subsequently, various forms of cover versions and musical aspects that cover versions share are defined. The thesis also deals in detail with the creation and distribution of a database of cover versions. Furthermore, the work presents methods and techniques for comparing and processing the calculated parameters. Attention is then paid to the OTI method, CSM calculation and methods dealing with parameter selection. The next part of the thesis is devoted to the design of systems for recognizing cover versions. Then there are compared systems already designed for recognizing cover versions. Furthermore, the thesis describes machine learning techniques and evaluation methods for evaluating the classification with a special emphasis on artificial neural networks. The last part of the thesis deals with the implementation of two systems in MATLAB and Python. These systems are then tested on the created database of cover versions.

Diploma thesis is aimed to trainable image segmentation using deep neural networks. In the paper is explained the principle of digital image processing and image segmentation. In the paper is also explained the principle of artificial neural network, model of artificial neuron, training and activation of artificial neural network. In practical part of the paper is created an algorithm of sliding window to generate sub-images from image from magnetic rezonance. Generated sub-images are used to train, test and validate of the model of neural network. In practical part of the paper si created the model of the artificial neural network, which is used to trainable image segmentation. Model of the neural network is created using the Deeplearning4j library and it is optimized to parallel training using Spark library.

Indiana University-Purdue University Indianapolis (IUPUI) / In recent years, deep learning models have become popular in the real-time embedded application, but there are many complexities for hardware deployment because of limited resources such as memory, computational power, and energy. Recent research in the field of deep learning focuses on reducing the model size of the Convolution Neural Network (CNN) by various compression techniques like Architectural compression, Pruning, Quantization, and Encoding (e.g., Huffman encoding). Network pruning is one of the promising technique to solve these problems. This thesis proposes methods to prune the convolution neural network (SqueezeNet) without introducing network sparsity in the pruned model. This thesis proposes three methods to prune the CNN to decrease the model size of CNN without a significant drop in the accuracy of the model. 1: Pruning based on Taylor expansion of change in cost function Delta C. 2: Pruning based on L2 normalization of activation maps. 3: Pruning based on a combination of method 1 and method 2. The proposed methods use various ranking methods to rank the convolution kernels and prune the lower ranked filters afterwards SqueezeNet model is fine-tuned by backpropagation. Transfer learning technique is used to train the SqueezeNet on the CIFAR-10 dataset. Results show that the proposed approach reduces the SqueezeNet model by 72% without a significant drop in the accuracy of the model (optimal pruning efficiency result). Results also show that Pruning based on a combination of Taylor expansion of the cost function and L2 normalization of activation maps achieves better pruning efficiency compared to other individual pruning criteria and most of the pruned kernels are from mid and high-level layers. The Pruned model is deployed on BlueBox 2.0 using RTMaps software and model performance was evaluated.

Artificial Inteligens and machine learning has started to get established as reco- gnizable terms to the general masses in their daily lives. Applications such as voice recognicion and image recognicion are used widely in mobile phones and autonomous systems such as self-drivning cars. This study examines how one can utilize this technique to classify sound as a complement to videosurveillan- ce in different settings, for example a busstation or other areas that might need monitoring. To be able to do this a technique called Convolution Neural Ne- twork has been used since this is a popular architecture to use when it comes to image classification. In this model every sound has a visual representation in form of a spectogram that showes frequencies over time. One of the main goals of this study has been to be able to apply this technique on so called IOT units to be able to classify sounds in real time, this because of the fact that these units are relativly affordable and requires little resources. A Rasberry Pi was used to run a prototype version using tensorflow &amp; keras as base api ́s. The studys re- sults show which parts that are important to consider to be able to get a good and reliable system, for example which hardware and software that is needed to get started. The results also shows what factors is important to be able to stream live sound and get reliable results, a classification models architecture is very important where different layers and parameters can have a large impact on the end result. / Termer som Artificiell Intelligens och maskininlärning har under de senaste åren börjat etablera sig hos den breda massan och är numera någonting som på- verkar nästan alla människors vardagliga liv i någon form. Vanliga använd- ningsområden är röststyrning och bildigenkänning som bland annat används i mobiltelefoner och autonoma system som självkörande bilar med mera. Den här studien utforskar hur man kan använda sig av denna teknik för att kunna klassi- ficera ljud som ett komplement till videoövervakning i olika miljöer, till exem- pel på en busstation eller andra övervakningsobjekt. För att göra detta har en teknik kallad Convolution Neural Network använts, vilket är en mycket populär arkitektur att använda vid klassificering av bilder. I denna modell har varje ljud fått en visuell representation i form av ett spektogram som visar frekvenser över tid. Ett av huvudmålen med denna studie har varit att kunna applicera denna teknik på så kallade IOT-enheter för att klassificera ljud i realtid. Dessa är rela- tivt billiga och resurssnåla enheter vilket gör dem till ett attraktivt alternativ för detta ändamål. I denna studie används en Raspberry Pi för att köra en prototyp- version med Tensorflow &amp; Keras som grund APIer. Studien visar bland annat på vilka moment och delar som är viktiga att tänka på för att få igång ett smidigt och pålitligt system, till exempel vilken hårdvara och mjukvara som krävs för att starta. Den visar också på vilka faktorer som spelar in för att kunna streama ljud med bra resultat, detta då en klassifikationsmodells arkitektur och upp- byggnad kan ha stor påverkan på slutresultatet.

The word “Perception” seems to be intuitive and maybe the most straightforward problem for the human brain because as a child we have been trained to classify images, detect objects, but for computers, it can be a daunting task. Giving intuition and reasoning to a computer which has mere capabilities to accept commands and process those commands is a big challenge. However, recent leaps in hardware development, sophisticated software frameworks, and mathematical techniques have made it a little less daunting if not easy. There are various applications built around to the concept of “Perception”. These applications require substantial computational resources, expensive hardware, and some sophisticated software frameworks. Building an application for perception for the embedded system is an entirely different ballgame. Embedded system is a culmination of hardware, software and peripherals developed for specific tasks with imposed constraints on memory and power. Therefore, the applications developed should keep in mind the memory and power constraints imposed due to the nature of these systems.Before 2012, the problems related to “Perception” such as classification, object detection were solved using algorithms with manually engineered features. However, in recent years, instead of manually engineering the features, these features are learned through learning algorithms. The game-changing architecture of Convolution Neural Networks proposed in 2012 by Alex K, provided a tremendous momentum in the direction of pushing Neural networks for perception. This thesis is an attempt to develop a convolution neural network architecture for embedded systems, i.e. an architecture that has a small model size and competitive accuracy. Recreate state-of-the-art architectures using fire module’s concept to reduce the model size of the architecture. The proposed compact models are feasible for deployment on embedded devices such as the Bluebox 2.0. Furthermore, attempts are made to integrate the compact Convolution Neural Network with object detection pipelines.