OEM&Lieferant Ausgabe 1/2020

93 ARRK Engineering www.arrk-engineering.com Website erator and one discriminator – to provide a basis for the framework and to augment the dataset. In this phase, the used dataset com- prised more than 1,000 photos of pedestrians. “Additional images were generated using the GAN to extend the dataset,” Diviš explains. “The GAN’s generator synthesized an image and the discriminator assessed the quality of this image. The interaction between these two neural networks allows us to extract the features from the original objects, generate new images and so extend the original data- set relatively easily.” Then, the classification network was trained on the original dataset and the test resultswere evaluated. To achieve the best possible results, ARRK used state-of- the-art architectures for all elements in the experiment. “The generalization of the object represents a challenge in image processing. The ba- sic question is: What defines a pedestrian?” says Diviš. “This can be easily answered by humans, since we generalize inductively. Neural networks, on the other hand, work deductively and require numerous examples to identify a specific object.” Furthermore is important to observe “corner cases” – special cases inwhich pedestrians are not recognized, because of a pedestrian’s unusual posture, an obtrusion blocking a sensor’s view, or poor lighting due to weather conditions. Data- sets typically lack suitable image material to classify these exceptional cases, but thanks to the GAN structure that has been estab- lished, ARRK has managed to supplement the dataset with computer-generated images and thus mitigate this problem. Optimization of object classification processes ARRK then beganwith comprehensive tests to gain a deeper understanding of the processes that underlie CNN training. In their analyses, experts looked at a number of processes that occur in neural networks and examined ap- proaches to understand the neurons’ flow of information. “Some neurons are more asso- ciated with the identification of pedestrians and produce stronger responses than others,” explains Diviš. “That’s why we’ve tested a range of scenarios in which we deactivated certain neurons to see how they influence decision-making processes. We could confirm that not every neuron responsible for identi- fying pedestrians needs to be activated, and in fact not removing some neurons can even lead to quicker and better results.” The framework that was created can be used to analyze these types of changes. All of this allows the stability of algorithms to be sustainably increased, which will serve to make autonomous driving safer. Precau- tions could be taken, for example, to reduce the risk of an “adversary attack”—the ex- ternal deployment of a malicious code dis- guised as a neutral image to compromise the neural network. This code generates a disturbance and influences the decisions of certain neurons, making it impossible to correctly recognize objects. The effects of these types of external disruptions could be reduced by removing inactive neurons, as this would provide fewer targets to attack in the neural network. “We will never be able to guarantee correct object classification 100 percent of the time,” says Diviš. “In the au- tomotive industry, our job is to identify and better understand vulnerabilities in neural networks. Only by doing so can we take ef- ficient counteractive measures and ensure maximum safety.” A system’s object clas- sification capabilities can also be improved immensely through the evaluation and com- bination of various data collected by sensors such as cameras, lidar, and radar. To train the neural networks in this framework, the GAN’s generator used an existing external dataset comprising photos of pedestrians to create computer-generated images. The discriminator then distinguished between genuine and generated images. This interaction will improve the generator’s ability to produce realistic images over time. Generator Discriminator Real samples Fake image latent space 80 noise z 256 12 128 24 64 48 128 48 256 24 512 12 ”fake” ”real” Graphic: © ARRK Engineering As part of its research activities, ARRK Engineering has developed the foundations of a framework to better under- standhowCNNsrecognize images,andultimatelyto improve their ability to do so. Its aim is to increase the probability that the system will detect a pedestrian. Image: © ARRK Engineering