Face recognition approaches that are based on deep convolutional neural networks (CNN) have been dominating the field. The performance improvements they have provided in the so called in-the-wild datasets are significant, however, their performance under image quality degradations have not been assessed, yet. This is particularly important, since in real-world face recognition applications, images may contain various kinds of degradations due to motion blur, noise, compression artifacts, color distortions, and occlusion. In this work, we have addressed this problem and analyzed the influence of these image degradations on the performance of deep CNN-based face recognition approaches using the standard LFW closed-set identification protocol. We have evaluated three popular deep CNN models, namely, the AlexNet, VGG-Face, and GoogLeNet. Results have indicated that blur, noise, and occlusion cause a significant decrease in performance, while deep CNN models are found to be robust to distortions, such as color distortions and change in color balance.

No abstract

According to new regulations, railway employees require computer-based training before they get a driver license. Training with simulators is now a must to increase the safety and efficiency of railway transportation. In this study, a train driving simulator with full cab and motion platform, called TRENSIM, was developed for Turkish State Railways. It was realized that constructing a data model of the railway was one of the most important aspects of the project. Many modules of TRENSIM make use of this data model in real time. We developed a general railway data model, called OpenRailway, expressed in Extensible Markup Language for logical description of railway networks. OpenRailway includes a railway model, which consists of signals, railway elements (catenaries, switches, etc.), road properties, surroundings near the railway road and their parameters. OpenRailway can also be modified to be used in the development of other vehicle training simulators. Data from real trains was collected in 4 ms periods for the validation of the dynamic and visualization models (about 200 GB). It was seen that data generated by models are appropriate to the collected real data.

No abstract