Object detection, a critical feature for autonomous vehicles, is performed today using Convolutional Neural Networks (CNNs). Errors in a CNN execution can modify the way the vehicle sense the surrounding environment, potentially causing accidents or unexpected behaviors. The high computational requirements of CNNs combined with the need to perform detection in real-time allow little margin for implementing error detection. In this project, an extremely efficient error detection solution for radiation induced errors in CNN is presented based on the observation that, in the absence of errors, the differences between the input frames and the detection provided by the CNN should be strictly correlated. In other words, if the image between two subsequent frames does not change significantly, the detection should also be very similar. Similarly, if the detection varies considerably from a frame to the next, then the input image should also have been different. Whenever input images and output detection don't correlate, it is possible to detect an error. After formalizing and evaluating the inter-frame and output correlation thresholds, the detection strategy is implemented and validated, utilizing data from previous radiation experiments. Exploiting the intrinsic efficiency in processing images of devices used to execute CNNs, up to 80% of errors are detected, while adding low overhead. The same error detection solution is then proposed to detect false positives in fault-free CNN executions. This strategy is also implemented and validated, utilizing ground truth annotations and fault-free CNN executions. For this application, 9% of the false positives can be detected reliably. A deeper analysis shows that more false positives can be detected, if a certain percentage of wrong detections is accepted.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.