Amir Nazemi

The National Highway Traffic Safety Administration reported that more than 90% of in-road accidents in 2015 occurred purely because of drivers? errors and misjudgments, with such factors as fatigue and other sorts of distractions being the main cause of these accidents [1]. One promising solution for reducing (or even resolving) such human errors is via autonomous or computer-assisted driving systems. Autonomous vehicles (AVs) are currently being designed with the aim of reducing fatalities in accidents by being insusceptible to typical driver errors. Moreover, in addition to improved safety, autonomous systems offer many other potential benefits to society: 1) improved fuel efficiency beyond that of human driving, making driving more cost beneficial and environmentally friendly, 2) reduced commute times due to improved driving behaviors and coordination among AVs, and 3) a better driving experience for individuals with disabilities, to name a few.

Vehicle Make and Model Recognition (MMR) systems provide a fully automatic framework to recognize and classify different vehicle models. Several approaches have been proposed to address this challenge, however they can perform in restricted conditions. Here, we formulate the vehicle make and model recognition as a fine-grained classification problem and propose a new configurable on-road vehicle make and model recognition framework. We benefit from the unsupervised feature learning methods and in more details we employ Locality constraint Linear Coding (LLC) method as a fast feature encoder for encoding the input SIFT features. The proposed method can perform in real environments of different conditions. This framework can recognize fifty models of vehicles and has an advantage to classify every other vehicle not belonging to one of the specified fifty classes as an unknown vehicle. The proposed MMR framework can be configured to become faster or more accurate based on the applicati...

Deep convolutional neural networks can be highly vulnerable to small perturbations of their inputs, potentially a major issue or limitation on system robustness when using deep networks as classifiers. In this paper we propose a low-cost method to explore marginal sample data near trained classifier decision boundaries, thus identifying potential adversarial samples. By finding such adversarial samples it is possible to reduce the search space of adversarial attack algorithms while keeping a reasonable successful perturbation rate. In our developed strategy, the potential adversarial samples represent only 61% of the test data, but in fact cover more than 82% of the adversarial samples produced by iFGSM and 92% of the adversarial samples successfully perturbed by DeepFool on CIFAR10.

ABSTRACT Automatic vehicle Make and Model Recognition (MMR) system offers a competent way to vehicle classification and recognition systems. This paper proposes a real time while robust vehicle make and model recognition system extracting the vehicle sub-image from the background and studies some sparse feature coding methods such as Orthogonal Matching Pursuit (OMP), some variation of Sparse Coding (SC) methods and compares them to choose the best one. Our method employs the sparse feature coding methods on dense Scale-Invariant Feature Transform (SIFT) features and Support Vector Machine (SVM) for classification. The proposed system is examined by an Iranian on road vehicles dataset, which its samples are in different point of views, various weather conditions and illuminations.

In this paper we propose a deep model for perceptual image en-hancement based on generative modeling. The proposed frame-work is inspired by the Conditional Variational AutoEncoder (CVAE)which is a well-known deep generative structure. In generativemodels, there are efficient regularizers for controlling the outputdistributions using information from input data which lead to accu-rate and visually plausible results with few parameters. Additionally,we propose to use an image quality assessment network to deter-mine the best result among those obtained by the implementedCVAEs. The proposed CVAE structure models the histogram vec-tors of different color channels and parameters of image data (i.e.,the networks do not work directly on pixel values). This configu-ration makes the proposed framework capable of using images ofdifferent sizes. Qualitative and numerical evaluations on a relateddataset compared to state-of-the-art indicate superiority of the pro-posed framework in improving i...