Computer Vision News - June 2023

8 CVPR Best Paper Award Candidate The model begins using a feature network to process each detected feature track separately. This network consists of existing common, fully convolutional layers, a correlation layer, and a recurrent layer . The correlation layer allows for correlating a grayscale image patch with event sequences, while the recurrent layer utilizes temporal information in the events, which inherently contain motion. “ Then we add a frame attention module on incremental flow at every step . Many feature trackers currently in the wild are based on optical flow. To make the model more robust to the different sorts of geometric changes that a patch might undergo at each step, we take the patch and apply similarity transform – rotating it, scaling it, and then forcing the network to predict the augmented flow. We found this strategy to be effective in closing the sim- to-real gap. ”