TY - JOUR
T1 - Seeing is worse than believing
T2 - 13th European Conference on Computer Vision, ECCV 2014
AU - Barbu, Andrei
AU - Barrett, Daniel P.
AU - Chen, Wei
AU - Siddharth, Narayanaswamy
AU - Xiong, Caiming
AU - Corso, Jason J.
AU - Fellbaum, Christiane D.
AU - Hanson, Catherine
AU - Hanson, Stephen José
AU - Hélie, Sébastien
AU - Malaia, Evguenia
AU - Pearlmutter, Barak A.
AU - Siskind, Jeffrey Mark
AU - Talavage, Thomas Michael
AU - Wilbur, Ronnie B.
PY - 2014
Y1 - 2014
N2 - We had human subjects perform a one-out-of-six class action recognition task from video stimuli while undergoing functional magnetic resonance imaging (fMRI). Support-vector machines (SVMs) were trained on the recovered brain scans to classify actions observed during imaging, yielding average classification accuracy of 69.73% when tested on scans from the same subject and of 34.80% when tested on scans from different subjects. An apples-to-apples comparison was performed with all publicly available software that implements state-of-the-art action recognition on the same video corpus with the same cross-validation regimen and same partitioning into training and test sets, yielding classification accuracies between 31.25% and 52.34%. This indicates that one can read people's minds better than state-of-the-art computer-vision methods can perform action recognition.
AB - We had human subjects perform a one-out-of-six class action recognition task from video stimuli while undergoing functional magnetic resonance imaging (fMRI). Support-vector machines (SVMs) were trained on the recovered brain scans to classify actions observed during imaging, yielding average classification accuracy of 69.73% when tested on scans from the same subject and of 34.80% when tested on scans from different subjects. An apples-to-apples comparison was performed with all publicly available software that implements state-of-the-art action recognition on the same video corpus with the same cross-validation regimen and same partitioning into training and test sets, yielding classification accuracies between 31.25% and 52.34%. This indicates that one can read people's minds better than state-of-the-art computer-vision methods can perform action recognition.
KW - action recognition
KW - fMRI
UR - http://www.scopus.com/inward/record.url?scp=84906500200&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84906500200&partnerID=8YFLogxK
U2 - 10.1007/978-3-319-10602-1_40
DO - 10.1007/978-3-319-10602-1_40
M3 - Conference article
AN - SCOPUS:84906500200
SN - 0302-9743
VL - 8693 LNCS
SP - 612
EP - 627
JO - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
JF - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
IS - PART 5
Y2 - 6 September 2014 through 12 September 2014
ER -