Figure-ground segmentation of natural scenes

This repository contains high precision manual figure-ground annotations and category labels for a set of frame sequences and static images of natural scenes. These annotations were used to analyze figure-ground motion and depth statistics in the manuscript: Statistical regularities in natural scenes that support figure-ground segregation by neuronal populationsClara T. Friedman, Minqi Wang, Thomas Yerxa, Bryce A. Arseneau, Xin Huang, and Emily A. CooperPLOS Computational Biology The annotations are separated into 2 datasets: Motion and Distance. The Motion dataset was collected new as part of this project and then video frames were annotated. The Distance dataset contains selected scenes from the UT Austin Natural Image Database. These scenes were selected because they have co-registered range measurements (see README in DataSets folder for more info). This repository contains the following folders stored as zip files: - Annotations: Segmentation masks for all frames and images. - DataSets: The raw frames in the Motion dataset, and instructions and metadata for downloading the Distance dataset from UT Austin.- Visuals: Visualizations of the annotations superimposed on the images, distance maps, and optic flow computed from the frames.- InterimFiles: Saved results from interim analysis steps in the associated Github code repository. These files can be used to skip over time-intensive processes steps and to use the original optic flows computed from raw video frames (instead of the anonmized frames included herein). Analysis code can be found at https://github.com/eacooper/FigureGround