This site contains description of and links to the benchmark
data
sets in the forms of synchronized multi-camera videos and time-stamped
RFID readings recorded and
annotated by the Laboratory for Assisted Cognition Environments (LACE)
at University of Rochester.
The data sets are generated for the Department of Defense MURI
project.
Currently, five participants are involved in the data collection in two
separate sessions, respectively.
In total, 10 data sets ( =
5 [subjects] x 2
[sessions] ) were collected in the
forms of videos and RFID readings.
Each of the 10 data sets contains 5 trials of the
following 6
activities performed in a continuous manner as follows:
Trials
trial-1
trial-2
..................
Activities
A01
A02
A03
A04
A05
A06
A01
A02
A03
A04
A05
A06
............
The RFID
labels-based annotation for the temporal segmentation of the activities
is downloadable here.
The individual activities are described below:
Activity Descriptions
Activity
Description
A01
Walking
around the indoor
space
A02
Sitting and watching TV
A03
Preparing
utensil (at
kitchen)
A04
Preparing
cereal (with
utensil and food)
A05
Drinking
water (with a cup)
A06
Storing
utensil (to cupboard)
Data Synchronization
The video file and RFID file in a give data set are synchronized as
follows:
Each
frame of a given AVI file contains four synchronized camera
views
(i.e., cam-1 in top left through cam-4 in bottom-right).
The RFID
file contains frame number, RFID label, and the associated
objet
label, which are commar-separatged according to the csv data format.
The list of the RFID tags attached on the various positions in the lab
is downloadable here.
AVI File Viewing Tip
Most of commercial /
free programs to view videos are time-based (i.e., the program
indicates the lapse time of the currently viewing frame in
terms
of seconds or miliseconds.) VirtualDub,
a free software to view AVI files, is frame-based (i.e., the program
indicates the lapse time of the currently viewing frame in terms of the
frame number.)
I recommend the use of VirtualDub to control the view
of individual frames.