This is for potential Computer Vision features. In principle this could be used for enhancing the AI/dialog capabilities of the bot, by adding visual context. The new OpenAI CLIP model is a promissing candidate that may lead to useful features. However, at the moment I don't have specific ideas in mind, except just fun/experimental stuff (like "see what the AI thinks you look like lol").
At the technical level this shouldnt be too difficult, as it would just mean feeding a camera frame to a CV neural net, perhaps when a certain command is issued.
This is for potential Computer Vision features. In principle this could be used for enhancing the AI/dialog capabilities of the bot, by adding visual context. The new OpenAI CLIP model is a promissing candidate that may lead to useful features. However, at the moment I don't have specific ideas in mind, except just fun/experimental stuff (like "see what the AI thinks you look like lol").
At the technical level this shouldnt be too difficult, as it would just mean feeding a camera frame to a CV neural net, perhaps when a certain command is issued.