How did you create the conditional image
Hi,
Thanks for sharing the amazing dataset, i saw a article from hugging which shared how to create dataset for image captioning and condition image formation. Do you mind sharing the library or. if if possible the code.
Looking forward to hearing from you.
Thanks,
Ankush Singal
Hello
If you have some problems with adding the second image to the dataset sample, you can do it using:
dataset = load_dataset("celeba_captioned", data_dir='./', features=datasets.Features({'image': datasets.Image(), 'text': datasets.Value(dtype='string'), 'conditioning_image': datasets.Image()}))
In this example your dataset is locally stored in ./celeba-captioned.
Of course you should also provide metadata.csvfile with all paths
i can also send you a .ipynb file, if you send me your email
Thank you, my email is [email protected]
Thanks , can you please share the library or code to form conditional images?. For image captioning i am using BLIP for image captioning and depth estimation for conditional image formation. But could not do on the whole dataset. Any help is appreciated.
I used the HED detector (you can find it in controlnet github), but actually i haven't succeed at training controlnet with this conditioning images
Seems like it overfits to this conditionings
May be I've done something wrong
Thank you very much, this is really helpful. Have a good day!!