Hello Kagglers,
\nI found a peculiar image (ID1963715583) showing a distinct, regular pattern of glowing red dots on the ground.
Identification Query: Does anyone recognize what these red dots are? (e.g., Laser projection, sensor artifact, or natural feature?)
\nModel Concern: I suspect this regular, unnatural pattern could negatively impact our model's accuracy and generalization in few sample datasets like this competition.
\nShould we filter/remove these images, or keep them?\nThanks!
", "messages": [ { "id": 3376976, "postDate": "2025-12-15T13:14:28.303Z", "content": "@qiyuliaocsiro May I ask whether the test set will include images like this?Thank you！！
", "rawMarkdown": "@qiyuliaocsiro May I ask whether the test set will include images like this?Thank you！！", "votes": 1, "replies": [ { "id": 3377027, "postDate": "2025-12-15T15:09:12.753Z", "content": "You may ask but pretty sure they may not reply :)
\nI see three red dot images in the train set. Looks like the same camera. Its not real hard to have one of the LLM's fix the red dots. Than run a test. Train and submit with and without the dots. Guess you could do a third test were you drop the 3 images.
\nThat will give you a clue - but remember that the current test score does not include all the private images, but you would have a bit more information to base any action decision on.
", "rawMarkdown": "You may ask but pretty sure they may not reply :)\n\nI see three red dot images in the train set. Looks like the same camera. Its not real hard to have one of the LLM's fix the red dots. Than run a test. Train and submit with and without the dots. Guess you could do a third test were you drop the 3 images.\n\nThat will give you a clue - but remember that the current test score does not include all the private images, but you would have a bit more information to base any action decision on.\n" } ] }, { "id": 3370027, "postDate": "2025-12-10T14:28:59.940Z", "content": "There are a few images with similar patterns, varying lighting conditions so varying effects. We are all playing with the same data, and we all run the risks of leader board probing, but, try photoshop them out, or remove them entirely, or leave them, and see what the LB shows. We do not know what the test set holds…
\nYou could also try and add them as an augment to other samples so that the algorithm learns to ignore them.
", "rawMarkdown": "There are a few images with similar patterns, varying lighting conditions so varying effects. We are all playing with the same data, and we all run the risks of leader board probing, but, try photoshop them out, or remove them entirely, or leave them, and see what the LB shows. We do not know what the test set holds... \n\nYou could also try and add them as an augment to other samples so that the algorithm learns to ignore them.", "votes": 2 }, { "id": 3369741, "postDate": "2025-12-10T09:54:56.347Z", "content": "Hello Kagglers,
\nI found a peculiar image (ID1963715583) showing a distinct, regular pattern of glowing red dots on the ground.
Identification Query: Does anyone recognize what these red dots are? (e.g., Laser projection, sensor artifact, or natural feature?)
\nModel Concern: I suspect this regular, unnatural pattern could negatively impact our model's accuracy and generalization in few sample datasets like this competition.
\nShould we filter/remove these images, or keep them?\nThanks!
", "rawMarkdown": "\nHello Kagglers, \n\nI found a peculiar image (ID1963715583) showing a distinct, regular pattern of glowing red dots on the ground.\n \nIdentification Query: Does anyone recognize what these red dots are? (e.g., Laser projection, sensor artifact, or natural feature?)\n \nModel Concern: I suspect this regular, unnatural pattern could negatively impact our model's accuracy and generalization in few sample datasets like this competition.\n \nShould we filter/remove these images, or keep them?\n\nThanks!", "votes": 2 }, { "id": 3370002, "postDate": "2025-12-10T14:11:19.317Z", "content": "These patterns could be in the test data also, so it is difficult to say anything @nakagawaren0805
", "rawMarkdown": "These patterns could be in the test data also, so it is difficult to say anything @nakagawaren0805 " }, { "id": 3369765, "postDate": "2025-12-10T10:20:22.317Z", "rawMarkdown": "", "votes": -1, "isDeleted": true } ], "comments": [ { "id": 3376976, "author_name": "Aic", "author_url": "", "post_date": "2025-12-15T13:14:28.303000", "content": "@qiyuliaocsiro May I ask whether the test set will include images like this?Thank you！！
", "votes": 1, "replies": [ { "id": 3377027, "author_name": "PC Jimmmy", "author_url": "", "post_date": "2025-12-15T15:09:12.753000", "content": "You may ask but pretty sure they may not reply :)
\nI see three red dot images in the train set. Looks like the same camera. Its not real hard to have one of the LLM's fix the red dots. Than run a test. Train and submit with and without the dots. Guess you could do a third test were you drop the 3 images.
\nThat will give you a clue - but remember that the current test score does not include all the private images, but you would have a bit more information to base any action decision on.
", "votes": 0, "replies": [] } ] }, { "id": 3370027, "author_name": "Tim Krige", "author_url": "", "post_date": "2025-12-10T14:28:59.940000", "content": "There are a few images with similar patterns, varying lighting conditions so varying effects. We are all playing with the same data, and we all run the risks of leader board probing, but, try photoshop them out, or remove them entirely, or leave them, and see what the LB shows. We do not know what the test set holds…
\nYou could also try and add them as an augment to other samples so that the algorithm learns to ignore them.
", "votes": 2, "replies": [] }, { "id": 3370002, "author_name": "Ravi Ramakrishnan", "author_url": "", "post_date": "2025-12-10T14:11:19.317000", "content": "These patterns could be in the test data also, so it is difficult to say anything @nakagawaren0805
", "votes": 0, "replies": [] }, { "id": 3369765, "author_name": "", "author_url": "", "post_date": "2025-12-10T10:20:22.317000", "content": "", "votes": -1, "replies": [] } ], "raw_markdown_by_id": { "3376976": "@qiyuliaocsiro May I ask whether the test set will include images like this?Thank you！！", "3370027": "There are a few images with similar patterns, varying lighting conditions so varying effects. We are all playing with the same data, and we all run the risks of leader board probing, but, try photoshop them out, or remove them entirely, or leave them, and see what the LB shows. We do not know what the test set holds... \n\nYou could also try and add them as an augment to other samples so that the algorithm learns to ignore them.", "3369741": "\nHello Kagglers, \n\nI found a peculiar image (ID1963715583) showing a distinct, regular pattern of glowing red dots on the ground.\n \nIdentification Query: Does anyone recognize what these red dots are? (e.g., Laser projection, sensor artifact, or natural feature?)\n \nModel Concern: I suspect this regular, unnatural pattern could negatively impact our model's accuracy and generalization in few sample datasets like this competition.\n \nShould we filter/remove these images, or keep them?\n\nThanks!", "3370002": "These patterns could be in the test data also, so it is difficult to say anything @nakagawaren0805 ", "3369765": "" } }