![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10846564%2F9991dff61b7bf153608ef2ab551cecae%2Fbad.png?generation=1682238975154392&alt=media\")
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10846564%2Fd82a3e8ba513a9891664c86582486996%2Fbad.png?generation=1682238979819480&alt=media\")
Good catch, thank you! This is a tricky area due to the border of the papyrus fragment having several physical layers in view. This particular case is interesting and I think has some room for ambiguity, but I agree that the regions you highlighted, though they are ink in the infrared photo, are not part of the segmented surface. I've updated the Frag2 ink label on my end to make that small change, and another organizer will likely post it soon. There was another small spot near the bottom right of the fragment with the same effect, which I updated also.
\nThese were clear enough for me to make a change to my own label images. In general though, there will be many small cases where the correct labeling is ambiguous or unclear. Feel free in these cases to experiment with making your own changes to the ink label, changing the surface mask to instruct your model not to sample those regions, or other techniques to build tolerance into your model.
", "rawMarkdown": "Good catch, thank you! This is a tricky area due to the border of the papyrus fragment having several physical layers in view. This particular case is interesting and I think has *some* room for ambiguity, but I agree that the regions you highlighted, though they are ink in the infrared photo, are not part of the segmented surface. I've updated the Frag2 ink label on my end to make that small change, and another organizer will likely post it soon. There was another small spot near the bottom right of the fragment with the same effect, which I updated also.\n\nThese were clear enough for me to make a change to my own label images. In general though, there will be many small cases where the correct labeling is ambiguous or unclear. Feel free in these cases to experiment with making your own changes to the ink label, changing the surface mask to instruct your model not to sample those regions, or other techniques to build tolerance into your model.", "votes": 4, "replies": [ { "id": 2248040, "postDate": "2023-05-06T13:13:03.200Z", "content": "Thank you for your reply 😄
", "rawMarkdown": "Thank you for your reply 😄" } ] }, { "id": 2231373, "postDate": "2023-04-23T08:51:21.990Z", "content": "", "rawMarkdown": "https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10846564%2Fc7c160ef94b53f4d689114816279c54a%2Fir(bad).jpg?generation=1682239722542206&alt=media", "votes": 1, "replies": [ { "id": 2240717, "postDate": "2023-04-30T19:26:56.607Z", "content": "May I ask where you found this image? Or did you plot it out yourself somehow?
", "rawMarkdown": "May I ask where you found this image? Or did you plot it out yourself somehow?", "replies": [ { "id": 2248047, "postDate": "2023-05-06T13:14:29.137Z", "content": "I applied a label mask to each photo and then just observed👀
", "rawMarkdown": "I applied a label mask to each photo and then just observed👀" } ] } ] }, { "id": 2231359, "postDate": "2023-04-23T08:45:13.257Z", "rawMarkdown": "", "isDeleted": true }, { "id": 2231357, "postDate": "2023-04-23T08:43:41.383Z", "rawMarkdown": "", "isDeleted": true } ], "comments": [ { "id": 2234847, "author_name": "Stephen Parsons", "author_url": "", "post_date": "2023-04-25T14:23:59.520000", "content": "Good catch, thank you! This is a tricky area due to the border of the papyrus fragment having several physical layers in view. This particular case is interesting and I think has some room for ambiguity, but I agree that the regions you highlighted, though they are ink in the infrared photo, are not part of the segmented surface. I've updated the Frag2 ink label on my end to make that small change, and another organizer will likely post it soon. There was another small spot near the bottom right of the fragment with the same effect, which I updated also.
\nThese were clear enough for me to make a change to my own label images. In general though, there will be many small cases where the correct labeling is ambiguous or unclear. Feel free in these cases to experiment with making your own changes to the ink label, changing the surface mask to instruct your model not to sample those regions, or other techniques to build tolerance into your model.
", "votes": 4, "replies": [ { "id": 2248040, "author_name": "Mecndo", "author_url": "", "post_date": "2023-05-06T13:13:03.200000", "content": "Thank you for your reply 😄
", "votes": 0, "replies": [] } ] }, { "id": 2231373, "author_name": "Mecndo", "author_url": "", "post_date": "2023-04-23T08:51:21.990000", "content": "", "votes": 1, "replies": [ { "id": 2240717, "author_name": "Chen Lin", "author_url": "", "post_date": "2023-04-30T19:26:56.607000", "content": "May I ask where you found this image? Or did you plot it out yourself somehow?
", "votes": 0, "replies": [ { "id": 2248047, "author_name": "Mecndo", "author_url": "", "post_date": "2023-05-06T13:14:29.137000", "content": "I applied a label mask to each photo and then just observed👀
", "votes": 0, "replies": [] } ] } ] }, { "id": 2231359, "author_name": "", "author_url": "", "post_date": "2023-04-23T08:45:13.257000", "content": "", "votes": 0, "replies": [] }, { "id": 2231357, "author_name": "", "author_url": "", "post_date": "2023-04-23T08:43:41.383000", "content": "", "votes": 0, "replies": [] } ], "raw_markdown_by_id": { "2231337": "", "2234847": "Good catch, thank you! This is a tricky area due to the border of the papyrus fragment having several physical layers in view. This particular case is interesting and I think has *some* room for ambiguity, but I agree that the regions you highlighted, though they are ink in the infrared photo, are not part of the segmented surface. I've updated the Frag2 ink label on my end to make that small change, and another organizer will likely post it soon. There was another small spot near the bottom right of the fragment with the same effect, which I updated also.\n\nThese were clear enough for me to make a change to my own label images. In general though, there will be many small cases where the correct labeling is ambiguous or unclear. Feel free in these cases to experiment with making your own changes to the ink label, changing the surface mask to instruct your model not to sample those regions, or other techniques to build tolerance into your model.", "2231373": "https://www.googleapis.com/download/storage/v1/b/kaggle-forum-message-attachments/o/inbox%2F10846564%2Fc7c160ef94b53f4d689114816279c54a%2Fir(bad).jpg?generation=1682239722542206&alt=media", "2231359": "", "2231357": "" } }