Access to training Data

[mr-sarthakgupta]

In the paper, it is mentioned that one could generate the training data using captions from images. Would it be possible to provide the training data used to train the models. I'd like to try and improve upon this work.

Also, please let us know the procedure for collection of test data, specifically the images listed in synthwildx.csv

[davin11]

• Next weeK, I will provide the captions list of the training.

• The images of the synthwildx dataset were handmade collected. In particular, we looked for images on the social network using the name of the generator as a tag. We also checked that there were no tags from other generators in the image description.

[ydvaayan]

Please provide the above listed resources. Also could you tell which text to image generator was used to generate the synthetic images in the reference set ?