I can't understand if there is a regulation about the intended method to use. \nSo a fine-tuning of powerful pre-trained models is against the rules or is accepted?
", "messages": [ { "id": "567977", "postDate": "07/04/2019 08:03:03", "content": "I can't understand if there is a regulation about the intended method to use. \nSo a fine-tuning of powerful pre-trained models is against the rules or is accepted?
", "rawMarkdown": "I can't understand if there is a regulation about the intended method to use. \nSo a fine-tuning of powerful pre-trained models is against the rules or is accepted?", "votes": null }, { "id": "576971", "postDate": "07/16/2019 08:05:38", "content": "It is accepted but you will need to disclose the source of the pre-trained models here
\n\nYou cannot use pre-trained models that are trained on the data, that is not available for other participants (for example, your company's internal models pre-trained on company's private datasets). However, you can use the models pre-trained on publicly available datasets even if they are not published. You will need to disclose the datasets used for pre-training in the Official External Data Thread
", "rawMarkdown": "It is accepted but you will need to disclose the source of the pre-trained models [here](https://www.kaggle.com/c/open-images-2019-visual-relationship/discussion/94290#latest-542389) \n\nYou cannot use pre-trained models that are trained on the data, that is not available for other participants (for example, your company's internal models pre-trained on company's private datasets). However, you can use the models pre-trained on publicly available datasets even if they are not published. You will need to disclose the datasets used for pre-training in the [Official External Data Thread](https://www.kaggle.com/c/open-images-2019-visual-relationship/discussion/94290#latest-542389)", "votes": null } ], "comments": [ { "id": 576971, "author_name": "akuznetsa", "author_url": "", "post_date": "07/16/2019 08:05:38", "content": "It is accepted but you will need to disclose the source of the pre-trained models here
\n\nYou cannot use pre-trained models that are trained on the data, that is not available for other participants (for example, your company's internal models pre-trained on company's private datasets). However, you can use the models pre-trained on publicly available datasets even if they are not published. You will need to disclose the datasets used for pre-training in the Official External Data Thread
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "567977": "I can't understand if there is a regulation about the intended method to use. \nSo a fine-tuning of powerful pre-trained models is against the rules or is accepted?", "576971": "It is accepted but you will need to disclose the source of the pre-trained models [here](https://www.kaggle.com/c/open-images-2019-visual-relationship/discussion/94290#latest-542389) \n\nYou cannot use pre-trained models that are trained on the data, that is not available for other participants (for example, your company's internal models pre-trained on company's private datasets). However, you can use the models pre-trained on publicly available datasets even if they are not published. You will need to disclose the datasets used for pre-training in the [Official External Data Thread](https://www.kaggle.com/c/open-images-2019-visual-relationship/discussion/94290#latest-542389)" }, "source": "meta" }