I have seen in various discussion threads that the rules of this competition mean that external data that is not allowed for commercial use cannot be used?
\nThen ImageNet pre-trained models cannot be used given that ImageNet's license is open for non-commercial research use only?
Is the interpretation that weights that are not available for commercial use should not be used due to this rule below?
\n\n\nOpen Source: You hereby license and will license your winning Submission and the source code used to generate the Submission under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code.
\n
To be honest, as with other competitions, I have been using ImageNet's pre-trained model without concern, but I am now concerned after reading the posts to the following discussion.
\nReference:
\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/375791#2085843
\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/373211#2071928
I have seen in various discussion threads that the rules of this competition mean that external data that is not allowed for commercial use cannot be used?
\nThen ImageNet pre-trained models cannot be used given that ImageNet's license is open for non-commercial research use only?
Is the interpretation that weights that are not available for commercial use should not be used due to this rule below?
\n\n\nOpen Source: You hereby license and will license your winning Submission and the source code used to generate the Submission under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code.
\n
To be honest, as with other competitions, I have been using ImageNet's pre-trained model without concern, but I am now concerned after reading the posts to the following discussion.
\nReference:
\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/375791#2085843
\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/373211#2071928
If the model has been licensed accordingly, they can be used. This applies to most ImageNet models.
\nBut some pre-trained models, like some ConvNextv2 weights/code, have specific licensing which disallows commercial usage.
\nJust my point of view.
", "rawMarkdown": "If the model has been licensed accordingly, they can be used. This applies to most ImageNet models.\n\nBut some pre-trained models, like some ConvNextv2 weights/code, have specific licensing which disallows commercial usage. \n\nJust my point of view.", "votes": 2, "replies": [ { "id": 2099674, "postDate": "2023-01-14T16:22:06.783Z", "content": "Thank you for your comment!
\nAs you told me, convnext v2 does mention that about the license, which I don't see on the convnext v1 github.
\nSince ImageNet itself is a dataset that is not available for commercial use, I was concerned that weights pre-trained with ImageNet could not be used.
\nOn the other hand, if ImageNet weights are usable, it might be interpreted that even if some mammography datasets which are not available for commercial-use, they can be used as long as they are publicly available.
A few questions remain, but my concerns have been alleviated, thank you!
", "rawMarkdown": "Thank you for your comment!\n\nAs you told me, convnext v2 does mention that about the license, which I don't see on the convnext v1 github.\n\nSince ImageNet itself is a dataset that is not available for commercial use, I was concerned that weights pre-trained with ImageNet could not be used.\nOn the other hand, if ImageNet weights are usable, it might be interpreted that even if some mammography datasets which are not available for commercial-use, they can be used as long as they are publicly available.\n\nA few questions remain, but my concerns have been alleviated, thank you!", "votes": 2, "replies": [ { "id": 2099700, "postDate": "2023-01-14T16:54:04.850Z", "content": "The whole licensing thing is a very troublesome topic in the field of ML.
", "rawMarkdown": "The whole licensing thing is a very troublesome topic in the field of ML. ", "votes": 3 } ] } ] } ], "comments": [ { "id": 2099409, "author_name": "Psi", "author_url": "", "post_date": "2023-01-14T12:19:34.823000", "content": "If the model has been licensed accordingly, they can be used. This applies to most ImageNet models.
\nBut some pre-trained models, like some ConvNextv2 weights/code, have specific licensing which disallows commercial usage.
\nJust my point of view.
", "votes": 2, "replies": [ { "id": 2099674, "author_name": "YYama", "author_url": "", "post_date": "2023-01-14T16:22:06.783000", "content": "Thank you for your comment!
\nAs you told me, convnext v2 does mention that about the license, which I don't see on the convnext v1 github.
\nSince ImageNet itself is a dataset that is not available for commercial use, I was concerned that weights pre-trained with ImageNet could not be used.
\nOn the other hand, if ImageNet weights are usable, it might be interpreted that even if some mammography datasets which are not available for commercial-use, they can be used as long as they are publicly available.
A few questions remain, but my concerns have been alleviated, thank you!
", "votes": 2, "replies": [ { "id": 2099700, "author_name": "Psi", "author_url": "", "post_date": "2023-01-14T16:54:04.850000", "content": "The whole licensing thing is a very troublesome topic in the field of ML.
", "votes": 3, "replies": [] } ] } ] } ], "raw_markdown_by_id": { "2099402": "I have seen in various discussion threads that the rules of this competition mean that external data that is not allowed for commercial use cannot be used?\nThen ImageNet pre-trained models cannot be used given that ImageNet's license is open for non-commercial research use only?\n\nIs the interpretation that weights that are not available for commercial use should not be used due to this rule below?\n\n>Open Source: You hereby license and will license your winning Submission and the source code used to generate the Submission under an Open Source Initiative-approved license (see www.opensource.org) that in no event limits commercial use of such code or model containing or depending on such code.\n\nTo be honest, as with other competitions, I have been using ImageNet's pre-trained model without concern, but I am now concerned after reading the posts to the following discussion.\n\nReference:\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/375791#2085843\nhttps://www.kaggle.com/competitions/rsna-breast-cancer-detection/discussion/373211#2071928\n", "2099409": "If the model has been licensed accordingly, they can be used. This applies to most ImageNet models.\n\nBut some pre-trained models, like some ConvNextv2 weights/code, have specific licensing which disallows commercial usage. \n\nJust my point of view." } }