Autonomous Vehicles need precise information as to the Drive-able space in order to be able to safely navigate. In recent years deep learning and Semantic Segmentation have attracted intense research. It is a highly advancing and rapidly evolving field that continues to provide excellent results. Research has shown that deep learning is emerging as a powerful tool in many applications. The aim of this study is to develop a deep learning system to estimate the Free Drive-able space. Building on the state of the art deep learning techniques, semantic segmentation will be used to replace the need for highly accurate maps, that are expensive to license. Free Drive-able space is defined as the drive-able space on the correct side of the road, that can be reached without a collision with another road user or pedestrian. A state of the art deep network will be trained with a custom data-set in order to learn complex driving decisions. Motivated by good results, further deep learning techniques will be applied to measure distance from monocular images. The findings demonstrate the power of deep learning techniques in complex driving decisions. The results also indicate the economic and technical feasibility of semantic segmentation over expensive high definition maps.

<p>In this dissertation, we document a few of our recent attempts in bridging the gap between the fast evolving deep learning research and the vast industry needs for dealing with computer vision challenges. More specifically, we developed novel deep-learning-based techniques for the following application-driven computer vision challenges: image super-resolution with quality restoration, motion estimation by optical flow, object detection for shape reconstruction, and object segmentation for motion tracking. Those four topics cover the computer vision hierarchy from the low level where digital images are processed to restore missing information for better human perception, to middle level where certain objects of interest are recognized and their motions are analyzed, finally to high level where the scene captured in the video footage will be interpreted for further analysis. In the process of building the whole-package of  ready-to-deploy solutions, we center our efforts on designing and training the most suitable convolutional neural networks for the particular computer vision problem at hand. Complementary procedures for data collection, data annotation,  post-processing of network outputs tailored for specific application needs, and deployment details will also be discussed where necessary. We hope our work demonstrates the applicability and versatility of convolutional neural networks for real-world computer vision tasks on a broad spectrum, from seeing better to understanding better.</p>

Training autonomous agents that are capable of performing their assigned job without fail is the ultimate goal of deep reinforcement learning. This thesis introduces a dueling Quantile Regression Deep Q-network, where the network learns the state value quantile function and advantage quantile function separately. With this network architecture the agent is able to learn to control simulated robots in the Gazebo simulator. Carefully crafted reward functions and state spaces must be designed for the agent to learn in complex non-stationary environments. When trained for only 100,000 timesteps, the agent is able reach asymptotic performance in environments with moving and stationary obstacles using only the data from the inertial measurement unit, LIDAR, and positional information. Through the use of transfer learning, the agents are also capable of formation control and flocking patterns. The performance of agents with frozen networks is improved through advice giving in Deep Q-networks by use of normalized Q-values and majority voting.

Tor is a Privacy-Enhancing Technology based on onion routing which lets its users browse the web anonymously. Even though the traffic is encrypted in multiple layers, traffic analysis can still be used to gather information from meta-data such as time, size, and direction of the traffic. A Website Fingerprinting (WF) attack is characterized by monitoring traffic locally to the user in order to predict the destination website based on the observed patterns. TrafficSliver is a proposed defence against WF attacks which splits the traffic on multiple paths in the Tor network. This way, a local attacker is assumed to only be able to observe a subset of all the user's total traffic. The initial evaluation of TrafficSliver against Deep Fingerprinting (DF), the state-of-the-art WF attack, showed promising results for the defence, reducing the accuracy of DF from over 98% down to less than 7% without adding artificial delays or dummy traffic. In this thesis, we further evaluate TrafficSliver against DF beyond what was done in the original work by De la Cadena et al. by using a richer data representation and finding out whether it is possible to utilize simulated training data to improve the accuracy of the attack. By introducing directional time as a richer data representation and increasing the size of the training dataset using a simulator, the accuracy of DF was improved against TrafficSliver on three different datasets. Against the original dataset provided by the authors of TrafficSliver, the accuracy was initially 7.1% and then improved to 49.9%. The results were confirmed by using two additional datasets with TrafficSliver, where the accuracy was improved from 5.4% to 44.9% and from 9.8% to 37.7%. / Tor är ett personlig-integritetsverktyg baserat på onion routing som låter sina användare anonymnt besöka hemsidor på internet. Även om trafiken är enkrypterad i flera lager, kan trafikanalys användas för att utvinna information från metadata som exempelvis: tid, storlek och riktning av trafik. En Website Fingerprinting (WF)-attack karaktäriseras av att övervaka trafik nära användaren för att sedan avgöra vilken hemsida som besökts utifrån mönster. TrafficSliver är ett föreslaget försvar mot WF-attacker genom att dela upp trafiken på flera vägar genom nätverket. Detta gör att en attackerare antas endast kunna se en delmängd av användarens totala trafik. Den första utvärderingen av TrafficSliver mot Deep Fingerprinting (DF), spjutspetsen inom WF-attacker, visade lovande resultat för försvaret genom att reducera träffsäkerheten av DF från över 98% till mindre än 7% utan att lägga till artificiella fördröjningar eller falsk trafik. I denna uppsats strävar vi att fortsätta utvärderingen av TrafficSliver mot DF utöver vad som redan har gjorts av De la Cadena et al. med en rikare datarepresentation och en undersökning huruvida det går att använda simulerad data för att träna attacker mot försvaret. Genom att introducera riktad tid och öka mängden data för att träna attacken, ökades träffsäkerheten av DF mot TrafficSliver på tre distinkta dataset. Mot det dataset som samlades in av TrafficSliver var träffsäkerheten inledelsevis 7.1% och sedan förbättrad med hjälp av riktad tid och större mängder av simulerad träningsdata till 49.9%. Dessa resultat bekräftades även för två ytterligare dataset med TrafficSliver, där träffsäkerheten blev förbättrad från 5.4% till 44.9% och från 9.8% till 37.7%.

No description available.

No description available.

No description available.

京都大学 / 新制・課程博士 / 博士(情報学) / 甲第24725号 / 情博第813号 / 新制||情||137(附属図書館) / 京都大学大学院情報学研究科知能情報学専攻 / (主査)教授 鹿島 久嗣, 教授 山本 章博, 教授 吉川 正俊 / 学位規則第4条第1項該当 / Doctor of Informatics / Kyoto University / DFAM

This dissertation presents novel methods for addressing important challenges related to the robustness of Deep Neural Networks (DNNs) for 3D understanding and in 3D setups. Our research focuses on two main areas, adversarial robustness on 3D data and setups and the robustness of DNNs to realistic 3D scenarios. One paradigm for 3D understanding is to represent 3D as a set of 3D points and learn functions on this set directly. Our first work, AdvPC, addresses the issue of limited transferability and ease of defense against current 3D point cloud adversarial attacks. By using a point cloud Auto-Encoder to generate more transferable attacks, AdvPC surpasses state-of-the-art attacks by a large margin on 3D point cloud attack transferability. Additionally, AdvPC increases the ability to break defenses by up to 38\% as compared to other baseline attacks on the ModelNet40 dataset. Another paradigm of 3D understanding is to perform 2D processing of multiple images of the 3D data. The second work, MVTN, addresses the problem of selecting viewpoints for 3D shape recognition using a Multi-View Transformation Network (MVTN) to learn optimal viewpoints. It combines MVTN with multi-view approaches leading to state-of-the-art results on standard benchmarks ModelNet40, ShapeNet Core55, and ScanObjectNN. MVTN also improves robustness to realistic scenarios like rotation and occlusion. Our third work analyzes the Semantic Robustness of 2D Deep Neural Networks, addressing the problem of high sensitivity toward semantic primitives in DNNs by visualizing the DNN global behavior as semantic maps and observing the interesting behavior of some DNNs. Additionally, we develop a bottom-up approach to detect robust regions of DNNs for scalable semantic robustness analysis and benchmarking of different DNNs. The fourth work, SADA, showcases the problem of lack of robustness in DNNs specifically for the safety-critical applications of autonomous navigation, beyond the simple classification setup. We present a general framework (BBGAN) for black-box adversarial attacks on trained agents, which covers semantic perturbations to the environment of the agent performing the task. BBGAN is trained to generate failure cases that consistently fool a trained agent on tasks such as object detection, self-driving, and autonomous UAV racing.

<p>Deep learning has been the driving force behind many successful data science achievements. However, the deep neural network (DNN) that forms the basis of deep learning is</p> <p>often over-parameterized, leading to training, prediction, and interpretation challenges. To</p> <p>address this issue, it is common practice to apply an appropriate penalty to each connection</p> <p>weight, limiting its magnitude. This approach is equivalent to imposing a prior distribution</p> <p>on each connection weight from a Bayesian perspective. This project offers a systematic investigation into the selection of the penalty function or prior distribution. Specifically, under</p> <p>the general theoretical framework of posterior consistency, we prove that consistent sparse</p> <p>deep learning can be achieved with a variety of penalty functions or prior distributions.</p> <p>Examples include amenable regularization penalties (such as MCP and SCAD), spike-and?slab priors (such as mixture Gaussian distribution and mixture Laplace distribution), and</p> <p>polynomial decayed priors (such as the student-t distribution). Our theory is supported by</p> <p>numerical results.</p> <p><br></p>