Context

My girlfriend’s master’s thesis was the Perception of Anthropomorphic Traits in Cars. She wanted to create a questionaire to test this hypothesis.

She had acquired a dataset of images of cars, and after selecting the images that could be used in the questionaire, she created a spreadsheet with the file name and the features for each car, such as the size of the grille, the shape of the headlights, etc.

She needed to select 10 images for the questionaire, and those images had to be representative of the different classes of the various labels, e.g. Bumper Shape: upturned lower edge-straight upper edge or Headlights Position: only upper.

Approach

To tackle this, I used scikit-multilearn to split the stimuli into train and test sets where all labels were represented, and used one of them for the questionaire.

You can find the notebook with the code here.