You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Thank you for your very interesting work! Unfortunately, I'm having trouble reproducing the zero-shot results on the RSNA Pneumonia dataset for both BioViL and BioViL-T. In the BioViL paper you write that you are using a test set of 9006 images, following the splits of GLoRIA. However, downloading the data from kaggle and using the GLoRIA splits, the test set has a size of 8006 images.
While this looks like a typo in the paper, I still cannot reproduce your results on these 8006 images. Can you confirm that the 8006 image test set following GLoRIA splits is indeed the test set you used or if it is not, which split was used in the end?
The text was updated successfully, but these errors were encountered:
Thank you for your very interesting work! Unfortunately, I'm having trouble reproducing the zero-shot results on the RSNA Pneumonia dataset for both BioViL and BioViL-T. In the BioViL paper you write that you are using a test set of 9006 images, following the splits of GLoRIA. However, downloading the data from kaggle and using the GLoRIA splits, the test set has a size of 8006 images.
While this looks like a typo in the paper, I still cannot reproduce your results on these 8006 images. Can you confirm that the 8006 image test set following GLoRIA splits is indeed the test set you used or if it is not, which split was used in the end?
The text was updated successfully, but these errors were encountered: