Question: ID Discrepancy Between part_data_22k Subset and Full 55k Dataset
Hello, thank you for this excellent work and for sharing this valuable dataset. I've been following the project for a while!
I was previously using the part_data_22k subset. Now that the full 55k dataset is available, I am trying to identify the original samples from the 22k subset within the new dataset by matching their IDs.
I expected to find a one-to-one match for most of the ~22k entries. However, my matching script only found 13,198 common IDs between the two sets.
Could you please clarify if this is the expected behavior? I wanted to check if perhaps some files were changed between versions, or if I might have an error in my matching process before I investigate my code further.
Thank you for your time and any insight you can provide!
Thank you for your interest in our work!
The difference in IDs is expected — during later stages, we performed additional processing on the original images, which caused the MD5-based IDs to change.
Therefore, it is normal that the matched IDs between the 22k subset and the 55k dataset do not fully align.
Thank you for your interest in our work!
The difference in IDs is expected — during later stages, we performed additional processing on the original images, which caused the MD5-based IDs to change.
Therefore, it is normal that the matched IDs between the 22k subset and the 55k dataset do not fully align.
Thanks for your reply!
Additionally, are the filtered 13198 identical ids files one-to-one correspondent between the 22k subset and 55k dataset?
Thank you for your follow-up question!
Yes — the 22k subset is a part of the full 55k dataset.
The 13,198 overlapping IDs you identified should correspond one-to-one between the two versions.
Thanks so much!!! Wish you sucess in your next research.