Background: Traffic Sign Recognition (TSR) is particularly useful for novice driversand self-driving cars. Driver Assistance Systems(DAS) involves automatic trafficsign recognition. Efficient classification of the traffic signs is required in DAS andunmanned vehicles for safe navigation. Convolutional Neural Networks(CNN) isknown for establishing promising results in the field of image classification, whichinspired us to employ this technique in our thesis. Computer vision is a process thatis used to understand the images and retrieve data from them. OpenCV is a Pythonlibrary used to detect traffic sign images in real-time. Objectives: This study deals with an experiment to build a CNN model which canclassify the traffic signs in real-time effectively using OpenCV. The model is builtwith low computational cost. The study also includes an experiment where variouscombinations of parameters are tuned to improve the model’s performance. Methods: The experimentation method involve building a CNN model based onmodified LeNet architecture with four convolutional layers, two max-pooling layersand two dense layers. The model is trained and tested with the German Traffic SignRecognition Benchmark (GTSRB) dataset. Parameter tuning with different combinationsof learning rate and epochs is done to improve the model’s performance.Later this model is used to classify the images introduced to the camera in real-time. Results: The graphs depicting the accuracy and loss of the model before and afterparameter tuning are presented. An experiment is done to classify the traffic signimage introduced to the camera by using the CNN model. High probability scoresare achieved during the process which is presented. Conclusions: The results show that the proposed model achieved 95% model accuracywith an optimum number of epochs, i.e., 30 and default optimum value oflearning rate, i.e., 0.001. High probabilities, i.e., above 75%, were achieved when themodel was tested using new real-time data.

Semantic Segmentation processes RGB scenes and classifies pixels collectively as an object. Recent deep learning methods have shown promising results in the accuracy and the speed of semantic segmentation. However, it is inevitable for the deep learning models to fall in overfitting to data used in training due to its nature of data-centric approaches. There have been numerous Regularization methods to overcome an overfitting problem, such as data augmentation, additional loss methods such as Euclidean or Least-Square terms, and structure-related methods by adding or modifying layers like Dropout and DropConnect in a network. Among those methods, penalizing a model via an additional loss or a weight constraint does not require memory increase. With this sight, our work purposes to improve a given segmentation model through temperatures and a lightweight discriminator. Temperatures have the role of generating different versions of probability maps through the division in softmax calculations. On top of probability maps from temperatures, we concatenate a simple discriminator after the segmentation network for the competition between groundtruth feature maps and modified feature maps. We pass the additional loss calculated from those probability maps into the principal network. Our contribution consists of two parts. Firstly, we use the adversarial loss as the regularization loss in the segmentation networks and validate that it can substitute the L2 regularization loss with better validation results. Also, we apply temperatures in segmentation probability maps for providing different information without using additional convolutional layers. The experiments indicate that the spiking temperature in a generator with keeping an original probability map in a discriminator provides the model improvement in terms of pixel accuracy and mean Intersection-of-Union (mIoU). Our framework shows that the segmentation model can be improved with a small increase in training time and the number of parameters.

Weld joint penetration determination is the key factor in welding process control area. Not only has it directly affected the weld joint mechanical properties, like fatigue for example. It also requires much of human intelligence, which either complex modeling or rich of welding experience. Therefore, weld penetration status identification has become the obstacle for intelligent welding system. In this dissertation, an innovative method has been proposed to detect the weld joint penetration status using machine-learning algorithms. A GTAW welding system is firstly built. Project a dot-structured laser pattern onto the weld pool surface during welding process, the reflected laser pattern is captured which contains all the information about the penetration status. An experienced welder is able to determine weld penetration status just based on the reflected laser pattern. However, it is difficult to characterize the images to extract key information that used to determine penetration status. To overcome the challenges in finding right features and accurately processing images to extract key features using conventional machine vision algorithms, we propose using convolutional neural network (CNN) to automatically extract key features and determine penetration status. Data-label pairs are needed to train a CNN. Therefore, an image acquiring system is designed to collect reflected laser pattern and the image of work-piece backside. Data augmentation is performed to enlarge the training data size, which resulting in 270,000 training data, 45,000 validation data and 45,000 test data. A six-layer convolutional neural network (CNN) has been designed and trained using a revised mini-batch gradient descent optimizer. Final test accuracy is 90.7% and using a voting mechanism based on three consequent images further improve the prediction accuracy.

Having enough data about the usage of tyre types on the road can lead to a better understanding of the consequences of studded tyres on the environment. This paper is focused on training and testing a machine learning model which can be further integrated into a larger system for automation of the data collection process. Different machine learning algorithms, namely CNN, SVM, and Random Forest, were compared in this experiment. The method used in this paper is an empirical method. First, sound data for studded and none-studded tyres was collected from three different locations in the city of Gävle/Sweden. A total of 760 Mel spectrograms from both classes was generated to train and test a well-known CNN model (AlexNet) on MATLAB. Sound features for both classes were extracted using JAudio to train and test models that use SVM and Random Forest classifi-ers on Weka. Unnecessary features were removed one by one from the list of features to improve the performance of the classifiers. The result shows that CNN achieved accuracy of 84%, SVM has the best performance both with and without removing some audio features (i.e 94% and 92%, respectively), while Random Forest has 89 % accuracy. The test data is comprised of 51% of the studded class and 49% of the none-studded class and the result of the SVM model has achieved more than 94 %. Therefore, it can be considered as an acceptable result that can be used in practice.

Abstract: Background: From Previous research, state-of-the-art deep neural networks have accomplished impressive results on many images classification tasks. However, adversarial attacks can easily fool these deep neural networks by adding little noise to the input images. This vulnerability causes a significant concern in deploying deep neural network-based systems in real-world security-sensitive situations. Therefore, research in attacking and the architectures with adversarial examples has drawn considerable attention. Here, we use the technique for image classification called Convolutional Neural Networks (CNN), which is known for determining favorable results in image classification. Objectives: This thesis reviews all types of adversarial attacks and CNN architectures in the present scientific literature. Experiment to build a CNN architecture to classify the handwritten digits in the MNIST dataset. And they are experimenting with adversarial attacks on the images to evaluate the accuracy fluctuations in categorizing images. This study also includes an experiment using the defensive distillation technique to improve the architecture's performance under adversarial attacks.  Methods: This thesis includes two methods; the systematic literature review method involved finding the best performing CNN architectures and best performing adversarial attack techniques. The experimentation method consists in building a CNN model based on modified LeNet architecture with two convolutional layers, one max-pooling layer, and two dropouts. The model is trained and tested with the MNIST dataset. Then applying adversarial attacks FGSM, IFGSM, MIFGSM on the input images to evaluate the model's performance. Later this model will be modified a little by defensive distillation technique and then tested towards adversarial attacks to evaluate the architecture's performance. Results: An experiment is conducted to evaluate the robustness of the CNN architecture in classifying the handwritten digits. The graphs show the accuracy before and after implementing adversarial attacks on the test dataset. The defensive distillation mechanism is applied to avoid adversarial attacks and achieve robust architecture. Conclusions: The results showed that FGSM, I-FGSM, MI-FGSM attacks reduce the test accuracy from 95% to around 35%. These three attacks to the proposed network successfully reduced ~70% of the test accuracy in all three cases for maximum epsilon 0.3. By the defensive distillation mechanism, the test accuracy reduces from 90% to 88% for max epsilon 0.3. The proposed defensive distillation process is successful in defending the adversarial attacks.

Aedes aegypti and Aedes albopictus are the main vector species for dengue disease and zika, two arboviruses that affect a substantial fraction of the global population. These mosquitoes breed in very slow-moving or standing pools of water, so detecting and managing these potential breeding habitats is a crucial step in preventing the spread of these diseases. Using high-resolution images collected by unmanned aerial vehicles (UAV) and their multispectral mapping data, this paper investigated bathymetry retrieval model in shallow water areas to help improve the habitat detection accuracy. While previous studies have found some success with shallow water bathymetry inversion on satellite imagery, accurate centimeter-level water depth regression from high-resolution, drone multispectral imagery still remains a challenge. Unlike previous retrieval methods generally relying on retrieval factor extraction and linear regression, this thesis introduced CNN methods, considering the nonlinear relationship between image pixel reflectance values and water depth. In order to look into CNN’s potential to retrieve shallow water depths from multispectral images captured by a drone, this thesis conducts a variety of case studies to respectively specify a proper CNN architecture, compare its performance in different datasets, band combinations, depth ranges and with other general bathymetry retrieval algorithms. In summary, the CNN-based model achieves the best regression accuracy of overall root mean square error lower than 0.5, in comparison with another machine learning algorithm, random forest, and 2 other semi-empirical methods, linear and ratio model, suggesting this thesis’s practical significance. / Aedes aegypti och Aedes albopictus är de viktigaste vektorarterna för dengue och zika, två arbovirus som drabbar en stor del av den globala befolkningen. Dessa myggor förökar sig i mycket långsamt rörliga eller stillastående vattensamlingar, så att upptäcka och hantera dessa potentiella förökningsmiljöer är ett avgörande steg för att förhindra spridningen av dessa sjukdomar. Med hjälp av högupplösta bilder som samlats in av obemannade flygfarkoster (UAV) och deras multispektrala kartläggningsdata undersöktes i den här artikeln en modell för att hämta batymetri i grunda vattenområden för att förbättra noggrannheten i upptäckten av livsmiljöer. Även om tidigare studier har haft viss framgång med inversion av bathymetri på grunt vatten med hjälp av satellitbilder, är det fortfarande en utmaning att göra en exakt regression av vattendjupet på centimeternivå från högupplösta, multispektrala bilder från drönare. Till skillnad från tidigare metoder som i allmänhet bygger på extrahering av återvinningsfaktorer och linjär regression, infördes i denna avhandling CNN-metoder som tar hänsyn till det icke-linjära förhållandet mellan bildpixlarnas reflektionsvärden och vattendjupet. För att undersöka CNN:s potential att hämta grunda vattendjup från multispektrala bilder som tagits av en drönare genomförs i denna avhandling en rad fallstudier för att specificera en lämplig CNN-arkitektur, jämföra dess prestanda i olika datamängder, bandkombinationer, djupintervall och med andra allmänna algoritmer för att hämta batymetri. Sammanfattningsvis uppnår den CNN-baserade modellen den bästa regressionsnoggrannheten med ett totalt medelkvadratfel som är lägre än 0,5, i jämförelse med en annan maskininlärningsalgoritm, random forest, och två andra halvempiriska metoder, linjär och kvotmodell, vilket tyder på den praktiska betydelsen av denna avhandling.

Gravel roads connect sparse populations and provide highways for agriculture and the transport of forest goods. Gravel roads are an economical choice where traffic volume is low. In Sweden, 21% of all public roads are state-owned gravel roads, covering over 20,200 km. In addition, there are some 74,000 km of gravel roads and 210,000 km of forest roads that are owned by the private sector. The Swedish Transport Administration (Trafikverket) rates the condition of gravel roads according to the severity of irregularities (e.g. corrugations and potholes), dust, loose gravel, and gravel cross-sections. This assessment is carried out during the summertime when roads are free of snow. One of the essential parameters for gravel road assessment is loose gravel. Loose gravel can cause a tire to slip, leading to a loss of driver control.  Assessment of gravel roads is carried out subjectively by taking images of road sections and adding some textual notes. A cost-effective, intelligent, and objective method for road assessment is lacking. Expensive methods, such as laser profiler trucks, are available and can offer road profiling with high accuracy. These methods are not applied to gravel roads, however, because of the need to maintain cost-efficiency.  In this thesis, we explored the idea that, in addition to machine vision, we could also use machine hearing to classify the condition of gravel roads in relation to loose gravel. Several suitable classical supervised learning and convolutional neural networks (CNN) were tested. When people drive on gravel roads, they can make sense of the road condition by listening to the gravel hitting the bottom of the car. The more we hear gravel hitting the bottom of the car, the more we can sense that there is a lot of loose gravel and, therefore, the road might be in a bad condition. Based on this idea, we hypothesized that machines could also undertake such a classification when trained with labeled sound data. Machines can identify gravel and non-gravel sounds. In this thesis, we used traditional machine learning algorithms, such as support vector machines (SVM), decision trees, and ensemble classification methods. We also explored CNN for classifying spectrograms of audio sounds and images in gravel roads. Both supervised learning and CNN were used, and results were compared for this study. In classical algorithms, when compared with other classifiers, ensemble bagged tree (EBT)-based classifiers performed best for classifying gravel and non-gravel sounds. EBT performance is also useful in reducing the misclassification of non-gravel sounds. The use of CNN also showed a 97.91% accuracy rate. Using CNN makes the classification process more intuitive because the network architecture takes responsibility for selecting the relevant training features. Furthermore, the classification results can be visualized on road maps, which can help road monitoring agencies assess road conditions and schedule maintenance activities for a particular road. / <p>Due to unforeseen circumstances the seminar was postponed from May 7 to 28, as duly stated in the new posting page.</p>

We investigate the novel idea of using computer vision to predict future stock price movement, which is performed by training a convolutional neural network (CNN) to detect patterns in images of stock graphs. Subsequently, we create a portfolio strategy based on the CNN stock price predictions to see if these predictions can generate alpha for investors. We apply this method in the Swedish stock market and evaluate the performance of CNN portfolios across two different exchanges and various stock indices segmented by market capitalisation. Our findings show that trading based on CNN predictions can outperform our benchmarks and generate positive alpha. Most of our portfolios generate positive alpha before transaction costs, while one also generates positive alpha after deducting transaction costs. Further, our results demonstrate that CNN models are capable of successfully generalising their trained knowledge, being able to detect information in stock graphs it has never seen before. This suggests that CNN models are not limited to the characteristics present in their training data, indicating that models trained under one set of market conditions can also be effective in a different market scenario. Our resultsfurther strengthen the overall findings of other researchers utilising similar methods as ours.

Manual proofreading of product packaging is a time-consuming and uncertain process that can pose significant challenges for companies, such as scalability issues, compliance risks and high costs. This thesis work introduces a novel solution by employing advanced computer vision and machine learning methods to automate the proofreading of multiple ingredients’ lists corresponding to multiple products simultaneously within a product package. By integrating Convolutional Neural Network (CNN) and Optical Character Recognition (OCR) techniques, this study examines the efficacy of automated proofreading in comparison to manual methods. The thesis involves analyzing product package artwork to identify ingredient lists utilizing the YOLOv5 object detection algorithm and the optical character recognition tool EasyOCR for ingredient extraction. Additionally, Python scripts are employed to extract ingredients from corresponding INCI PDF files (document that lists the standardized names of ingredients used in cosmetic products). A comprehensive comparison is then conducted to evaluate the accuracy and efficiency of automated proofreading. The comparison of the extracted ingredients from the product packages and their corresponding INCI PDF files yielded a match of 12.7%. Despite the suboptimal result, insights from the study highlights the limitations of current detection and recognition algorithms when applied to complex artwork. A few examples of the insights have been that the trained YOLOv5 model cuts through sentences in the ingredient list or that EasyOCR cannot extract ingredients from vertically aligned product package images. The findings underscore the need for advancements in detection algorithms and OCR tools to effectively handle objects like product packaging designs. The study also suggests that companies, such as H&amp;M, consider updating their artwork and INCI PDF files to align with the capabilities of current AI-driven tools. By doing so, they can enhance the efficiency and overall effectiveness of automated proofreading processes, thereby reducing errors and improving accuracy. / Manuell korrekturläsning av produktförpackningar är en tidskrävande och osäker process som kan skapa betydande utmaningar för företag, såsom skalbarhetsproblem, efterlevnadsrisker och höga kostnader. Detta examensarbete presenterar en ny lösning genom att använda avancerade metoder inom datorseende och maskininlärning för att automatisera korrekturläsningen av flera ingredienslistor som motsvarar flera produkter samtidigt inom en produktförpackning. Genom att integrera Convolutional Neural Network (CNN) och Optical Character Recognition (OCR) utreder denna studie effektiviteten av automatiserad korrekturläsning i jämförelse med manuella metoder. Avhandlingen analyserar designen av produktförpackningar för att identifiera ingredienslistor med hjälp av objektdetekteringsalgoritmen YOLOv5 och det optiska teckenigenkänningsverktyget EasyOCR för extrahera enskilda ingredienser från listorna. Utöver detta används Python-skript för att extrahera ingredienser från motsvarande INCI-PDF filer (dokument med standardiserade namn på ingredienser som används i kosmetika produkter). En omfattande jämförelse genomförs sedan för att utvärdera noggrannheten och effektiviteten hos automatiserad korrekturläsning. Jämförelsen av de extraherade ingredienserna från produktförpackningarna och deras korresponderande INCI-PDF filer gav ett matchnings resultat på 12.7%. Trots de mindre optimala resultaten belyser studien de begränsningar som finns hos de nuvarande detekterings- och teckenigenkänningsalgoritmerna när de appliceras på komplexa verk av produktförpackningar. Ett fåtal exempel på insikterna är bland annat att den tränade YOLOv5 modellen skär igenom meningar i ingredienslistan eller att EasyOCR inte kan extrahera ingredienser från stående ingredienslistor på produktförpackningsbilder. Resultaten understryker behovet av framsteg inom detekteringsalgoritmer och OCR-verktyg för att effektivt kunna hantera komplexa objekt som produktförpackningar. Studien föreslår även att företag, såsom H&amp;M, överväger att uppdatera sina design av produktförpackningar och INCI-PDF filer för att anpassa sig till kapaciteten hos aktuella AI-drivna verktyg. Genom att utföra detta kan de förbättra både effektiviteten och den övergripande kvaliteten hos de automatiserade korrekturläsningsprocesserna, vilket minskar fel och ökar noggrannheten.

Facial expression recognition (FER) is still a challenging concept, and machines struggle to comprehend effectively the dynamic shifts in facial expressions of human emotions. The existing systems, which have proven to be effective, consist of deeper network structures that need powerful and expensive hardware. The deeper the network is, the longer the training and the testing. Many systems use expensive GPUs to make the process faster. To remedy the above challenges while maintaining the main goal of improving the accuracy rate of the recognition, we create a generic hierarchical structure with variable settings. This generic structure has a hierarchy of three convolutional blocks, two dropout blocks and one fully connected block. From this generic structure we derived four different network structures to be investigated according to their performances. From each network structure case, we again derived six network structures in relation to the variable parameters. The variable parameters under analysis are the size of the filters of the convolutional maps and the max-pooling as well as the number of convolutional maps. In total, we have 24 network structures to investigate, and six network structures per case. After simulations, the results achieved after many repeated experiments showed in the group of case 1; case 1a emerged as the top performer of that group, and case 2a, case 3c and case 4c outperformed others in their respective groups. The comparison of the winners of the 4 groups indicates that case 2a is the optimal structure with optimal parameters; case 2a network structure outperformed other group winners. Considerations were done when choosing the best network structure, considerations were; minimum accuracy, average accuracy and maximum accuracy after 15 times of repeated training and analysis of results. All 24 proposed network structures were tested using two of the most used FER datasets, the CK+ and the JAFFE. After repeated simulations the results demonstrate that our inexpensive optimal network architecture achieved 98.11 % accuracy using the CK+ dataset. We also tested our optimal network architecture with the JAFFE dataset, the experimental results show 84.38 % by using just a standard CPU and easier procedures. We also compared the four group winners with other existing FER models performances recorded recently in two studies. These FER models used the same two datasets, the CK+ and the JAFFE. Three of our four group winners (case 1a, case 2a and case 4c) recorded only 1.22 % less than the accuracy of the top performer model when using the CK+ dataset, and two of our network structures, case 2a and case 3c came in third, beating other models when using the JAFFE dataset. / Electrical and Mining Engineering