 |
    |
Seeing is Worse than Believing: Reading People’s Minds Better than Computer-Vision Methods Recognize Actions
Andrei Barbu1, Daniel P. Barrett2, Wei Chen3, Narayanaswamy Siddharth4, Caiming Xiong5, Jason J. Corso6, Christiane D. Fellbaum7, Catherine Hanson8, Stephen José Hanson8, Sébastien Hélie2, Evguenia Malaia9, Barak A. Pearlmutter10, Jeffrey Mark Siskind2, Thomas Michael Talavage2, and Ronnie B. Wilbur2
1MIT, Cambridge, MA, USA
andrei@0xab.com
2Purdue University, West Lafayette, IN, USA
dpbarret@purdue.edu
shelie@purdue.edu
tmt@purdue.edu
wilbur@purdue.edu
3SUNY Buffalo, Buffalo, NY, USA
wchen23@buffalo.edu
4Stanford University, Stanford, CA, USA
nsid@stanford.edu
5University of California at Los Angeles, Los Angeles, CA, USA
caimingxiong@ucla.edu
6University of Michigan, Ann Arbor, MI, USA
jjcorso@eecs.umich.edu
7Princeton University, Princeton, NJ, USA
fellbaum@princeton.edu
8Rutgers University, Newark, NJ, USA
cat@psychology.rutgers.edu
jose@psychology.rutgers.edu
9University of Texas at Arlington, Arlington, TX, USA
malaia@uta.edu
10National University of Ireland Maynooth, Co. Kildare, Ireland
barak@cs.nuim.ie
Abstract. 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.
Keywords: action recognition, fMRI
LNCS 8693, p. 612 ff.
lncs@springer.com
© Springer International Publishing Switzerland 2014