ImageNet Roulette is a object recognition systems trained with one of the most popular object recognition training set ImageNet. In ImageNet there is a category for persons and this was used to train the system to classify people. When a user stands in front of the installation version of ImageNet Roulette the person is detected and classified. On a screen mirroring the person a green box with a label is overlaying the video footage. Â
In the online version a image was uploaded and all detected persons were boxed with the green frame with the label.
Under the image the labels linked to the categories and subcategories of WordNet the taxonomy used for labeling images in the ImageNet dataset.
(I am using the installation version for the situation)