3 ms·
I think more often it’s about translation. From image to classification ex. with no hope of mapping back.
by johanvts 2y ago
I think more often it’s about translation. From image to classification ex. with no hope of mapping back.
- the8472 2y agoThe classifier still has to learn a model, which is a lossy, compressed representation of the world. Classifiers can be very lossy. The outputs from Google's Deepdream are the kind of things that a classifier would "uncompress" to.