Back to Event Basics: Self-Supervised Learning of Image Reconstruction for Event Cameras via Photometric Constancy

Event cameras are novel vision sensors that sample, in an asynchronous fashion, brightness increments with low latency and high temporal resolution. The resulting streams of events are of high value by themselves, especially for high speed motion estimation. However, a growing body of work has also focused on the reconstruction of intensity frames from the events, as this allows bridging the gap with the existing literature on appearance-and frame-based computer vision. Recent work has mostly approached this problem using neural networks trained with synthetic, ground-truth data. In this work we approach, for the first time, the intensity reconstruction problem from a self-supervised learning perspective. Our method, which leverages the knowledge of the inner workings of event cameras, combines estimated optical flow and the event-based photometric constancy to train neural networks without the need for any ground-truth or synthetic data. Results across multiple datasets show that the performance of the proposed self-supervised approach is in line with the state-of-the-art. Additionally, we propose a novel, lightweight neural network for optical flow estimation that achieves high speed inference with only a minor drop in performance.

Recent work has mostly approached this problem from a machine learning perspective. With their E2VID artificial neural network, Rebecq et al. [3,4] were the first to show that learning-based methods trained to maximize perceptual similarity via supervised learning outperform hand-crafted techniques by a large margin in terms of image quality. Later, Scheerlinck et al. [5] achieved high speed inference with FireNet, a simplified model of E2VID. Despite the high levels of accuracy reported, these architectures were trained with large sets of synthetic data from event camera simulators [6], which adds extra complexity to the reconstruction problem due to the simulator-to-reality gap.