Awesome

Navigation Benchmark Scenarios

This repository contains standard scenarios for benchmarking navigation agent performance on the SUNCG and the Matterport3D datasets.

The scenarios are defined in CSV (comma separated value) format files where each row provides the state for one episode. The following column values are provided for each episode:

episodeId: integer id for given episode within the set.
split: train, val, or test indicating split for the episode.
task: string id for task type, one of p for point_goal, o for object_goal, or r for room_goal.
sceneId: id of scene within which episode takes place.
level: integer id of level (typically floor) within scene.
startX, startY, startZ: coordinates of agent starting position in scene space.
startAngle, startTilt: azimuth angle (counter-clockwise from scene space +X axis) and vertical tilt angles of agent starting state.
goalRoomId, goalRoomType: string id and category of room in which the goal resides. Required for room_goal, optional otherwise.
goalObjectId, goalObjectType: string id and category of goal object. Required for object_goal, optional otherwise.
goalX, goalY, goalZ: coordinates of goal point in scene space. Required for all task types. Position of goal for point_goal, position of object centroid for object_goal, or arbitrary point within goal room for room_goal.
goalAngle, goalTilt: angles defining desired view of goal. Reserved, currently not used.
dist, pathDist: Euclidean and geodesic (along shortest path) distance from agent start position to goal position
pathNumDoors, pathDoorIds: integer number of doors along shortest path from start to goal, and string of :-separated door ids along path. Note that doors at start or goal points are included.
pathNumRooms, pathRoomIndices: integer number of rooms along shortest path from start to goal and string of :-separated room ids along path. Note that starting and ending rooms are included.

Changelog

[0.1.0] - 2018-06-03

Initial beta release

Added v1 scenarios for SUNCG and Matterport3D