AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Is There a Way to Recognize Objects Robustly?
Deep learning can be used to recognize objects very robustly, especially in lots of partial information domains. The problem is that there's a lot of missing inform and we need to reason to fill in the gaps. But maybe molty modality is a way of overcoming that problem. And this is exactly the pitch you could use for grands because, and that's what you write in your introductions about mot modality. So it takes both imports and does things with them even though they don't look like each other. We have two modalities. It could be more. I meani ideally it should be as many as possible, because then you are super robust,. but you have vision