=================================


Leeds Robotic Commands Dataset


=================================










Contact Information: 


===========================



Author: Muhannad Alomari


Email1: scmara@leeds.ac.uk


Email2: omari.1988@gmail.com











Description:


===========================


Leeds Robotic Commands is a dataset of real-world RGB-D scenes of a robot manipulating different objects together with natural language descriptions of these actions. The scenes were recorded using a Microsoft Kinect2 sensor, and the descriptions were annotated by non-expert volunteers. The dataset includes 204 videos consisting of 17,373 frames in total. The dataset contains a total of 1024 commands, average of five-per video. A total of 51 different objects are manipulated in the videos such as basic block shapes, fruits, cutlery, and office supplies.











Contents:


===========================



1- Read_me.txt  : contains information about this dataset.



2- scenes 1-50.zip     : contains the first 50 scenes, scene1.zip to scene50.zip.



3- scenes 51-100.zip   : contains the second 50 scenes, scene51.zip to scene100.zip.



4- scenes 101-150.zip  : contains the third 50 scenes, scene101.zip to scene150.zip.



5- scenes 151-204.zip  : contains the last 54 scenes, scene151.zip to scene204.zip.



Each scene#.zip contains the data for a single scene, which is divded into 10 folders.



	a-  annotation    : contains a .txt file with all the natural language commands/annotations.

	b-  objects       : contains the pointclouds of the individual detected objects using the tabletop  object detection algorithm.

	c-  object_tracks : contains the 3d tracks of the detected objects in each frame.

	d-  rgb_cam       : contains the images from an external camera pointed at the robot (not used for object detection).

	e-  rgb_kinect    : contains the images from the Kinect2 fixed on the robot's chest (used to detect objects).

	f-  rgb_lefthand  : contains the images recorded from the robot's left arm camera.

	g-  rgb_righthand : contains the images recorded from the robot's right arm camera.

	h-  robot_state   : contains .txt files that recorded the robot joint angles and commands.

	i-  table_pointcloud    : contains the pointclouds of the detected table.

	j- tabletop_pointcloud : contains the pointclouds of all the objects detected on the table.

for more information on how to read/process pointclouds.

please refer to the point cloud library (http://pointclouds.org/).











Published papers:


===========================





@InProceedings{Alomari:2017,

 Title                    = {Natural Language Acquisition and Grounding for Embodied Robotic Systems},

 Author                   = {Alomari, M. and Duckworth, P. and Hogg, D. C. and Cohn, A. G.},

 Booktitle                = {Proc. of Association for the Advancement of Artificial Intelligence {(AAAI)}},

 Year                     = {2017}}