The paper "FlowNet: Learning Optical Flow with Convolutional Networks" by Philipp Fischer et al. addresses the challenge of using convolutional neural networks (CNNs) for optical flow estimation, a task that has not been successfully tackled by CNNs in the past. The authors propose two CNN architectures: FlowNetSimple and FlowNetCorr, with the latter including a correlation layer to aid in feature matching. To address the lack of large-scale ground truth datasets, they generate a synthetic Flying Chairs dataset, which consists of random background images from Flickr with segmented chair images overlayed. Despite the unrealistic nature of this dataset, the networks trained on it generalize well to real-world datasets like Sintel and KITTI, achieving competitive accuracy at frame rates of 5 to 10 fps. The paper also discusses related work, network architectures, training data, and experimental results, demonstrating the effectiveness of their approach in both synthetic and real-world scenarios.The paper "FlowNet: Learning Optical Flow with Convolutional Networks" by Philipp Fischer et al. addresses the challenge of using convolutional neural networks (CNNs) for optical flow estimation, a task that has not been successfully tackled by CNNs in the past. The authors propose two CNN architectures: FlowNetSimple and FlowNetCorr, with the latter including a correlation layer to aid in feature matching. To address the lack of large-scale ground truth datasets, they generate a synthetic Flying Chairs dataset, which consists of random background images from Flickr with segmented chair images overlayed. Despite the unrealistic nature of this dataset, the networks trained on it generalize well to real-world datasets like Sintel and KITTI, achieving competitive accuracy at frame rates of 5 to 10 fps. The paper also discusses related work, network architectures, training data, and experimental results, demonstrating the effectiveness of their approach in both synthetic and real-world scenarios.