Could you please share your experience about collecting data?
#4
by
thesby
- opened
Thank you for your great work of the caption model. I want to fine tune the model to give more details about cloth, but writing caption by me will cost a lot of time. Could you please share your experience about collecting caption? Is it possible to generate caption by a model or a workflow?
Having JoyCaption or another strong VLM like Qwen 2-VL write a set of captions, and then tweaking/fixing them by hand for your specific use case would the fastest way to do it.