Suppose i need to cluster images in to two groups and would like to cross validate by including equal number of images on each validation block. I had not done k fold cross validation with images before. Is there any kernel/or code references for that? Any suggestion please.
", "messages": [ { "id": "836047", "postDate": "05/06/2020 17:17:01", "content": "Suppose i need to cluster images in to two groups and would like to cross validate by including equal number of images on each validation block. I had not done k fold cross validation with images before. Is there any kernel/or code references for that? Any suggestion please.
", "rawMarkdown": "Suppose i need to cluster images in to two groups and would like to cross validate by including equal number of images on each validation block. I had not done k fold cross validation with images before. Is there any kernel/or code references for that? Any suggestion please.", "votes": null }, { "id": "836383", "postDate": "05/06/2020 23:58:26", "content": "Notebooks search show: K-fold CV, none with images till now. Maybe anyone from the Comp or even out it can give you a tip.
", "rawMarkdown": "Notebooks search show: K-fold CV, none with images till now. Maybe anyone from the Comp or even out it can give you a tip.", "votes": null }, { "id": "836472", "postDate": "05/07/2020 02:32:31", "content": "thanks @mpwolke .iam working on that.Looking for some documentation.but still i had not found a better one.
", "rawMarkdown": "thanks @mpwolke .iam working on that.Looking for some documentation.but still i had not found a better one.", "votes": null }, { "id": "837781", "postDate": "05/08/2020 03:31:27", "content": "Hi, maybe you're interested in https://www.kaggle.com/c/bengaliai-cv19/discussion/134929\nYou can find what you need in Sklearn lib.
\nStritifiedKFold is probable better than K-fold.
ok..thanks @bryce1010 Let me have a look through it.
", "rawMarkdown": "ok..thanks @bryce1010 Let me have a look through it.", "votes": null }, { "id": "838319", "postDate": "05/08/2020 13:30:12", "content": "This is not a regular image segmentation competition where you can do simple k-fold and average predictions like on any other data.
\n\nHere you will need in addition to do a proper ensembling of bounding boxes predictions
", "rawMarkdown": "This is not a regular image segmentation competition where you can do simple k-fold and average predictions like on any other data.\n\nHere you will need in addition to do a proper ensembling of bounding boxes predictions", "votes": null }, { "id": "838528", "postDate": "05/08/2020 16:18:17", "content": "Thank you for the advice Serigne.
", "rawMarkdown": "Thank you for the advice Serigne.", "votes": null }, { "id": "838561", "postDate": "05/08/2020 16:37:15", "content": "@Serigne thanks
", "rawMarkdown": "Serigne thanks", "votes": null } ], "comments": [ { "id": 836383, "author_name": "mpwolke", "author_url": "", "post_date": "05/06/2020 23:58:26", "content": "Notebooks search show: K-fold CV, none with images till now. Maybe anyone from the Comp or even out it can give you a tip.
", "votes": null, "replies": [ { "id": 836472, "author_name": "arunmohan003", "author_url": "", "post_date": "05/07/2020 02:32:31", "content": "thanks @mpwolke .iam working on that.Looking for some documentation.but still i had not found a better one.
", "votes": null, "replies": [] } ] }, { "id": 837781, "author_name": "bryce1010", "author_url": "", "post_date": "05/08/2020 03:31:27", "content": "Hi, maybe you're interested in https://www.kaggle.com/c/bengaliai-cv19/discussion/134929\nYou can find what you need in Sklearn lib.
\nStritifiedKFold is probable better than K-fold.
ok..thanks @bryce1010 Let me have a look through it.
", "votes": null, "replies": [] } ] }, { "id": 838319, "author_name": "serigne", "author_url": "", "post_date": "05/08/2020 13:30:12", "content": "This is not a regular image segmentation competition where you can do simple k-fold and average predictions like on any other data.
\n\nHere you will need in addition to do a proper ensembling of bounding boxes predictions
", "votes": null, "replies": [ { "id": 838528, "author_name": "mpwolke", "author_url": "", "post_date": "05/08/2020 16:18:17", "content": "Thank you for the advice Serigne.
", "votes": null, "replies": [] }, { "id": 838561, "author_name": "arunmohan003", "author_url": "", "post_date": "05/08/2020 16:37:15", "content": "@Serigne thanks
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "836047": "Suppose i need to cluster images in to two groups and would like to cross validate by including equal number of images on each validation block. I had not done k fold cross validation with images before. Is there any kernel/or code references for that? Any suggestion please.", "836383": "Notebooks search show: K-fold CV, none with images till now. Maybe anyone from the Comp or even out it can give you a tip.", "836472": "thanks @mpwolke .iam working on that.Looking for some documentation.but still i had not found a better one.", "837781": "Hi, maybe you're interested in https://www.kaggle.com/c/bengaliai-cv19/discussion/134929\nYou can find what you need in Sklearn lib. \nStritifiedKFold is probable better than K-fold.", "837941": "ok..thanks @bryce1010 Let me have a look through it.", "838319": "This is not a regular image segmentation competition where you can do simple k-fold and average predictions like on any other data.\n\nHere you will need in addition to do a proper ensembling of bounding boxes predictions", "838528": "Thank you for the advice Serigne.", "838561": "Serigne thanks" }, "source": "meta" }