Intensity-Image Reconstruction for Event Cameras Using Convolutional Neural Network

Abstract

Notice of Violation of IEEE Publication Principles <br><br> "Intensity-Image Reconstruction for Event Cameras Using Convolutional Neural Network," <br> by Yongwei Chen, Weitong Chen, Xixin Cao and Qianting Hua <br> in the Proceedings of the IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP), 2020, pp. 1973-1977 <br><br> After careful and considered review of the content and authorship of this paper by a duly constituted expert committee, this paper has been found to be in violation of IEEE’s Publication Principles. <br><br> This paper contains significant portions of original text from the paper cited below. The original text was copied without attribution (including appropriate references to the original author(s) and/or paper title) and without permission. <br><br> “Events-to-Video: Bringing Modern Computer Vision to Event Cameras”<br> by Henri Rebecq, René Ranftl, Vladlen Koltun, Davide Scaramuzza <br> in the Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Long Beach, 2019.<br><br> <br/> Event cameras have many benefits than conventional cameras, such as high temporal resolution, high dynamic range. However, because the outputs of event cameras are asynchronous event streams than intensity images, Frame-based algorithms cannot be directly used. It is also necessary to present intensity images of event cameras on the display for human viewing. In this paper, "event frames" are recovered from event streams in an attenuation method and they are fed into the U-net network to generate intensity images. Our model is trained on a large amount of simulated data and gradually reduces the perceptual loss through training. In order to evaluate the model, we compare the generated image with the target image on the simulated data and the real data. This proves that our model can reconstruct intensity images of event cameras very well.