I am using subsequences https://www.kaggle.com/julian3833/reef-a-cv-strategy-subsequences to divide the dataset. Use https://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302241 for F2 score statistics.
\nI used the weight of Big Brother Sheep https://www.kaggle.com/steamedsheep/yolov5-is-all-you-need in local cv0.503, using stacking https://www.kaggle.com/naturezhang/yolov5 -detections-tracking-on-cot has LB of 0.643. But the model I trained has 0.523 in local CV, but in C the LB is only 0.556.
\nSlowly the local CV is greater than 0.503 but why is the LB so different.
\nThis is very strange Can someone tell me why?
I am using subsequences https://www.kaggle.com/julian3833/reef-a-cv-strategy-subsequences to divide the dataset. Use https://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302241 for F2 score statistics.
\nI used the weight of Big Brother Sheep https://www.kaggle.com/steamedsheep/yolov5-is-all-you-need in local cv0.503, using stacking https://www.kaggle.com/naturezhang/yolov5 -detections-tracking-on-cot has LB of 0.643. But the model I trained has 0.523 in local CV, but in C the LB is only 0.556.
\nSlowly the local CV is greater than 0.503 but why is the LB so different.
\nThis is very strange Can someone tell me why?
helphelphelp
", "rawMarkdown": "helphelphelp", "votes": null }, { "id": "1670750", "postDate": "01/31/2022 20:44:38", "content": "Check out this discussion here:
\nhttps://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302057
It talks about how there is probably some training images in the LB. So while your CV is seeing new data the LB might be seeing new data + training data. It's worth thinking about.
", "rawMarkdown": "Check out this discussion here:\nhttps://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302057\n\nIt talks about how there is probably some training images in the LB. So while your CV is seeing new data the LB might be seeing new data + training data. It's worth thinking about.", "votes": null }, { "id": "1671040", "postDate": "02/01/2022 05:32:41", "content": "Thank you for your answer
", "rawMarkdown": "Thank you for your answer", "votes": null } ], "comments": [ { "id": 1670339, "author_name": "henini", "author_url": "", "post_date": "01/31/2022 12:44:07", "content": "helphelphelp
", "votes": null, "replies": [] }, { "id": 1670750, "author_name": "ivanaerlic", "author_url": "", "post_date": "01/31/2022 20:44:38", "content": "Check out this discussion here:
\nhttps://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302057
It talks about how there is probably some training images in the LB. So while your CV is seeing new data the LB might be seeing new data + training data. It's worth thinking about.
", "votes": null, "replies": [ { "id": 1671040, "author_name": "henini", "author_url": "", "post_date": "02/01/2022 05:32:41", "content": "Thank you for your answer
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1670262": "I am using subsequences [https://www.kaggle.com/julian3833/reef-a-cv-strategy-subsequences](url) to divide the dataset. Use [https://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302241](url) for F2 score statistics.\nI used the weight of Big Brother Sheep [https://www.kaggle.com/steamedsheep/yolov5-is-all-you-need](url) in local cv0.503, using stacking [https://www.kaggle.com/naturezhang/yolov5 -detections-tracking-on-cot](url) has LB of 0.643. But the model I trained has 0.523 in local CV, but in C the LB is only 0.556.\nSlowly the local CV is greater than 0.503 but why is the LB so different.\nThis is very strange Can someone tell me why?", "1670339": "helphelphelp", "1670750": "Check out this discussion here:\nhttps://www.kaggle.com/c/tensorflow-great-barrier-reef/discussion/302057\n\nIt talks about how there is probably some training images in the LB. So while your CV is seeing new data the LB might be seeing new data + training data. It's worth thinking about.", "1671040": "Thank you for your answer" }, "source": "meta" }