5 ms·
I agree that a manual option should have always been available. I'm just being pedantic: the context here is a "saliency algorithm." In the context of computer
by alanbernstein 5y ago
I agree that a manual option should have always been available. I'm just being pedantic: the context here is a "saliency algorithm." In the context of computer vision, the term saliency has a specific meaning. It's about visual perception of images, which doesn't generally include understanding letters, words, sentences, or humor.
- rwmj 5y agoCan't a neural network be trained so it recognizes cartoon panels and only shows the first ones? Seems at least plausible.
- alanbernstein 5y agoSure, if it's designed for cropping cartoon images to show the first panel. This algorithm is designed to work for any image shared on twitter. What should it do for cartoons where the last panel is a much more interesting visual image? Single-panel cartoons that include both the setup and the punchline in one frame (https://twitter.com/PunHubOnline/status/1392509471482683399/photo/1 https://twitter.com/PunHubOnline/status/1392509471482683399/...)? How about before-and-after comparison images?