AI-powered
podcast player
Listen to all your favourite podcasts with AI-powered features
Using the Internetet in a Training Process Is a Good Idea
Sketches are put into a network that has some sort of sementic meaning. If your sketch was a chair, then you want to train the network to give you a vector that corresponds to the word chair. You repeat that for hety shapes. In the case of tets shapes, what you can do is then take multipe random rendering of your hids shape. So now you have, lets say, 12 images for your tiden shape. And you go through that same process where you are in cutting them into a cementicaway descript and so because you've done that for each of those modalities, you now have it descripto for anything that you throw at.