This is my first research competition I am asking if we can use pretrained models?\nBecause I saw that the internet access is prohibited in this competition and pretrained models should be download via internet..
", "messages": [ { "id": "502707", "postDate": "03/29/2019 00:33:41", "content": "This is my first research competition I am asking if we can use pretrained models?\nBecause I saw that the internet access is prohibited in this competition and pretrained models should be download via internet..
", "rawMarkdown": "This is my first research competition I am asking if we can use pretrained models?\nBecause I saw that the internet access is prohibited in this competition and pretrained models should be download via internet..", "votes": null }, { "id": "502871", "postDate": "03/29/2019 07:11:26", "content": "You can use pretrained models published in Kaggle. \nFor example: pytorch pretrained models
", "rawMarkdown": "You can use pretrained models published in Kaggle. \nFor example: [pytorch pretrained models](https://www.kaggle.com/pvlima/pretrained-pytorch-models)", "votes": null }, { "id": "502886", "postDate": "03/29/2019 07:39:47", "content": "In this example i see \"additional files\" which disabled in this competition
", "rawMarkdown": "In this example i see \"additional files\" which disabled in this competition", "votes": null }, { "id": "502892", "postDate": "03/29/2019 07:52:04", "content": "I think in this case, you can create pretrained model dataset by yourself. Make it public and report to the organizer.
\nIn your kernel, you can add this dataset (pretrained) as well
This is the official statement from https://www.kaggle.com/c/imet-2019-fgvc6/rules
\n\n> Pretrained models may be used to construct the algorithms from publicly released academic datasets (ImageNet, iNaturalist, etc) full citation needed. Please specify any and all external data used for training when uploading results in the specified discussion post.
\n\nYou can publish your dataset including pretrained model you want to use. There is a good explanation how to upload and publish external data(pretrained model).\nhttps://www.kaggle.com/c/petfinder-adoption-prediction#Kernels-FAQ
", "rawMarkdown": "This is the official statement from https://www.kaggle.com/c/imet-2019-fgvc6/rules\n\n> Pretrained models may be used to construct the algorithms from publicly released academic datasets (ImageNet, iNaturalist, etc) full citation needed. Please specify any and all external data used for training when uploading results in the specified discussion post.\n\nYou can publish your dataset including pretrained model you want to use. There is a good explanation how to upload and publish external data(pretrained model).\nhttps://www.kaggle.com/c/petfinder-adoption-prediction#Kernels-FAQ", "votes": null }, { "id": "503161", "postDate": "03/29/2019 14:47:12", "content": "I suppose you dont need to create dataset by yourself but you can add existing one if there are models that you need...\nFor instance, I added two sets of models:
\n\nhttps://www.kaggle.com/pvlima/pretrained-pytorch-models\nhttps://www.kaggle.com/bminixhofer/pytorch-pretrained-image-models
\n\nAll models are from torchvision (docs: https://pytorch.org/docs/stable/torchvision/models.html ) and pretrained on ImageNet (Deng, J. et.al., ImageNet: A Large-Scale Hierarchical Image Database, CVPR09, 2009)
", "rawMarkdown": "I suppose you dont need to create dataset by yourself but you can add existing one if there are models that you need...\nFor instance, I added two sets of models:\n\nhttps://www.kaggle.com/pvlima/pretrained-pytorch-models\nhttps://www.kaggle.com/bminixhofer/pytorch-pretrained-image-models\n\nAll models are from torchvision (docs: https://pytorch.org/docs/stable/torchvision/models.html ) and pretrained on ImageNet (Deng, J. et.al., ImageNet: A Large-Scale Hierarchical Image Database, CVPR09, 2009)", "votes": null } ], "comments": [ { "id": 502871, "author_name": "backaggle", "author_url": "", "post_date": "03/29/2019 07:11:26", "content": "You can use pretrained models published in Kaggle. \nFor example: pytorch pretrained models
", "votes": null, "replies": [ { "id": 502886, "author_name": "stalkermustang", "author_url": "", "post_date": "03/29/2019 07:39:47", "content": "In this example i see \"additional files\" which disabled in this competition
", "votes": null, "replies": [] }, { "id": 502892, "author_name": "backaggle", "author_url": "", "post_date": "03/29/2019 07:52:04", "content": "I think in this case, you can create pretrained model dataset by yourself. Make it public and report to the organizer.
\nIn your kernel, you can add this dataset (pretrained) as well
This is the official statement from https://www.kaggle.com/c/imet-2019-fgvc6/rules
\n\n> Pretrained models may be used to construct the algorithms from publicly released academic datasets (ImageNet, iNaturalist, etc) full citation needed. Please specify any and all external data used for training when uploading results in the specified discussion post.
\n\nYou can publish your dataset including pretrained model you want to use. There is a good explanation how to upload and publish external data(pretrained model).\nhttps://www.kaggle.com/c/petfinder-adoption-prediction#Kernels-FAQ
", "votes": null, "replies": [] }, { "id": 503161, "author_name": "vlad0922", "author_url": "", "post_date": "03/29/2019 14:47:12", "content": "I suppose you dont need to create dataset by yourself but you can add existing one if there are models that you need...\nFor instance, I added two sets of models:
\n\nhttps://www.kaggle.com/pvlima/pretrained-pytorch-models\nhttps://www.kaggle.com/bminixhofer/pytorch-pretrained-image-models
\n\nAll models are from torchvision (docs: https://pytorch.org/docs/stable/torchvision/models.html ) and pretrained on ImageNet (Deng, J. et.al., ImageNet: A Large-Scale Hierarchical Image Database, CVPR09, 2009)
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "502707": "This is my first research competition I am asking if we can use pretrained models?\nBecause I saw that the internet access is prohibited in this competition and pretrained models should be download via internet..", "502871": "You can use pretrained models published in Kaggle. \nFor example: [pytorch pretrained models](https://www.kaggle.com/pvlima/pretrained-pytorch-models)", "502886": "In this example i see \"additional files\" which disabled in this competition", "502892": "I think in this case, you can create pretrained model dataset by yourself. Make it public and report to the organizer. \nIn your kernel, you can add this dataset (pretrained) as well", "502995": "This is the official statement from https://www.kaggle.com/c/imet-2019-fgvc6/rules\n\n> Pretrained models may be used to construct the algorithms from publicly released academic datasets (ImageNet, iNaturalist, etc) full citation needed. Please specify any and all external data used for training when uploading results in the specified discussion post.\n\nYou can publish your dataset including pretrained model you want to use. There is a good explanation how to upload and publish external data(pretrained model).\nhttps://www.kaggle.com/c/petfinder-adoption-prediction#Kernels-FAQ", "503161": "I suppose you dont need to create dataset by yourself but you can add existing one if there are models that you need...\nFor instance, I added two sets of models:\n\nhttps://www.kaggle.com/pvlima/pretrained-pytorch-models\nhttps://www.kaggle.com/bminixhofer/pytorch-pretrained-image-models\n\nAll models are from torchvision (docs: https://pytorch.org/docs/stable/torchvision/models.html ) and pretrained on ImageNet (Deng, J. et.al., ImageNet: A Large-Scale Hierarchical Image Database, CVPR09, 2009)" }, "source": "meta" }