4 ms·
It's not the same thing. For example, given a GUI with a titlebar, title, subtitle, text, and buttons, a human can instantly understand spatially the relationsh
by reasonableklout 3mo ago
It's not the same thing. For example, given a GUI with a titlebar, title, subtitle, text, and buttons, a human can instantly understand spatially the relationship between these items. But a naive OCR of such a GUI would be a flat stream of text that loses a ton of information.
- kortilla 3mo agoBut that’s not how models handle images either. They spatially segment and reason about title bars, placement, etc.