Text-to-Image
Diffusers
lora
flux
cape

Dataset caption

#2
by umairahmad1789 - opened

hey, thanks for sharing your work and results are very good. Can you share how you caption your dataset?

thank you! i used florence2 for the captioning, but also made some manual adjustments to highlight the details i wanted the model to learn

Thank you @martintomov for the response. I will try this in my next test for sure.

It would be very helpful for me to understand if you can share a sample caption?.
When captioning, I typically describe everything I want the model to learn, so I can modify it during inference. For instance, if I have ten images of a person and he is wearing glasses in three, I'll mention the glasses in those captions so that i can modify that during inference. What general rules do you follow for captioning?

@umairahmad1789 sorry for the late reply, i was super busy, and the notifications slipped under my radar. i decided to post the dataset, so here you go - hope it helps: martintomov/rayban-meta-glasses

Sign up or log in to comment