Closed Ph0eNiX0803 closed 2 years ago
Thank you so much for reporting the generation result. During our generation, we also faced the same problem. We solved this by removing the data without captions during post-processing. If you have enough computing resources and memory for your computer, you can consider the same post-processing with us. It could also be an alternative to add a function to delete the failed examples during dataset generation.
@Ph0eNiX0803 Hi, how many image pairs do you generate finally?
I ran the code from 0 to 3, it generated 4 sets images but got only 2 captions.