Artis Dobrajs

Monitoring, detection, and control of traffic is a serious problem in many cities and on roads around the world and poses a problem for effective and safe control and management of pedestrians with edge devices. Systems using the computer vision approach must ensure the safety of citizens and minimize the risk of traffic collisions. This approach is well suited for multiple object detection by automatic video surveillance cameras on roads, highways, and pedestrian walkways. A new Annotated Virtual Detection Line (AVDL) dataset is presented for multiple object detection, consisting of 74,108 data files and 74,108 manually annotated files divided into six classes: Vehicles, Trucks, Pedestrians, Bicycles, Motorcycles, and Scooters from the video. The data were captured from real road scenes using 50 video cameras from the leading video camera manufacturers at different road locations and under different meteorological conditions. The AVDL dataset consists of two directories, the Data directory and the Labels directory. Both directories provide the data as NumPy arrays. The dataset can be used to train and test deep neural network models for traffic and pedestrian detection, recognition, and counting.

Semantic segmentation of an incoming visual stream from cameras is an essential part of the perception system of self-driving cars. State-of-the-art results in semantic segmentation have been achieved with deep neural networks (DNNs), yet training them requires large datasets, which are difficult and costly to acquire and time-consuming to label. A viable alternative to training DNNs solely on real-world datasets is to augment them with synthetic images, which can be easily modified and generated in large numbers. In the present study, we aim at improving the accuracy of semantic segmentation of urban scenes by augmenting the Cityscapes real-world dataset with synthetic images generated with the open-source driving simulator CARLA (Car Learning to Act). Augmentation with synthetic images with a low degree of photorealism from the MICC-SRI (Media Integration and Communication Center–Semantic Road Inpainting) dataset does not result in the improvement of the accuracy of semantic segmentation, yet both MobileNetV2 and Xception DNNs used in the present study demonstrate a better accuracy after training on the custom-made CCM (Cityscapes-CARLA Mixed) dataset, which contains both real-world Cityscapes images and high-resolution synthetic images generated with CARLA, than after training only on the real-world Cityscapes images. However, the accuracy of semantic segmentation does not improve proportionally to the amount of the synthetic data used for augmentation, which indicates that augmentation with a larger amount of synthetic data is not always better.