3 ms·
Uhm. You’re basically asking how the entire NN works. There is no easy explanation for that.
by andsens 4y ago
Uhm. You’re basically asking how the entire NN works. There is no easy explanation for that.
- alok-g 4y agoI understand neural networks, embeddings, convolutions, etc. The part that's unclear to me is specifically how textual embeddings are linked into the img-to-img network trying to reduce the noise. In other words, am missing how the process is 'conditioned upon' the text. (I lack a understanding the same for conditional GANs as well.) If the answer is just that the textual embeddings are also fed as simple inputs to the network, I already understand then.
- capableweb 4y agoMight be worth looking through the dataset it was trained on, here's on example: https://laion-aesthetic.datasette.io/laion-aesthetic-6pls/images?_search=Bob+Moog&_sort=rowid https://laion-aesthetic.datasette.io/laion-aesthetic-6pls/im... So the model understands (kinda) who Bob Moog is, so when you include "Bob Moog" in the prompt, the model knows what you are looking for.
- ShamelessC 4y agoWhy did they unnecassarily re-index a smaller subset of Laion Aesthetic? You can search _all_ of laion using the pre-built faiss indices from laion.. https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2Fknn5.laion.ai&index=laion5B&useMclip=false https://rom1504.github.io/clip-retrieval/?back=https%3A%2F%2... is a hosted version, but you can download and host it yourself as well.