Psychophysical experiments with neural networks

Organization
Institute of Psychology
Abstract
Purpose of this project is to run some simple psychophysical experiments using pre-trained convolutional neural networks as “subjects”. You have to generate sets of simple images, use these images to train a neural network, and test its performance using test sets of images. The results should reveal similarities and differences between neural networks and human vision.
Graduation Theses defence year
2017-2018
Supervisor
Endel Põder
Spoken language (s)
Estonian, English
Requirements for candidates
Elementary knowledge in image processing, neural networks, and software for that purpose (for example, Matlab).
Level
Keywords

Application of contact

 
Name
Endel Põder
Phone
E-mail
endel.poder@ut.ee