We really admire your brilliant work! And we want to query about a single dataset used in your experiment.
In Table 3, you have mentioned that datasets highlighted in grey was not used in training process. However, the v7w dataset appeared in the 363k dataset json file. So we wonder whether you have used the v7w dataset in finetuning or not.
What's more, we find the Visual7w dataset is storaged with parquet pattern and it is difficult to recover those images. We can recover them to jpg pattern, but the name of those images can not match those written in the 363k dataset json file. Can you tell us how to get those data files in proper pattern? We are really looking forward to your reply! Wish you a great day!
We really admire your brilliant work! And we want to query about a single dataset used in your experiment.
In Table 3, you have mentioned that datasets highlighted in grey was not used in training process. However, the v7w dataset appeared in the 363k dataset json file. So we wonder whether you have used the v7w dataset in finetuning or not.
What's more, we find the Visual7w dataset is storaged with parquet pattern and it is difficult to recover those images. We can recover them to jpg pattern, but the name of those images can not match those written in the 363k dataset json file. Can you tell us how to get those data files in proper pattern? We are really looking forward to your reply! Wish you a great day!