3 ms·
DeepMind Shows AI Has Trouble Seeing Homer Simpson's Actions
- pavement 9y agoOh come on, this is an easy one. They need to include the complete audio channel in their training model, and identify important audio cues. 99% of Homer's characterization is really Dan Castellaneta's voice. The rest is simply the basic stretch-and-squash rubber band concept, as applied to the bug-eyed, yellow Simpsons visual character model. https://en.wikipedia.org/wiki/Rubber_hose_animation https://en.wikipedia.org/wiki/Rubber_hose_animation This is not an image processing task.