{ "id": 305901, "title": "About additional annotations of training images", "url": "/competitions/happy-whale-and-dolphin/discussion/305901", "author_name": "", "post_date": "2022-02-07T12:38:43.252365300Z", "votes": 6, "comment_count": 3, "views": 0, "content": "

Hello! My question is about of bounding box annotations of training images.
\nIn the past https://www.kaggle.com/c/humpback-whale-identification/ competition in rules(COMPETITION-SPECIFIC RULES, EXTERNAL DATA section) the following was pointed out:

\n
\n

Entrants may re-annotate images in the training set, but may not hand-label predictions, including having human observers rate and evaluate the test data set.

\n
\n

But in this competition I haven't found the same notes about additional(re-labeling) annotations.
\nCan I do bounding bbox annotations of subset of training(only training subset, not test) images or not?
\nIt's not clear for me.
\nThanks.

", "messages": [ { "id": "1679838", "postDate": "02/07/2022 12:38:43", "content": "

Hello! My question is about of bounding box annotations of training images.
\nIn the past https://www.kaggle.com/c/humpback-whale-identification/ competition in rules(COMPETITION-SPECIFIC RULES, EXTERNAL DATA section) the following was pointed out:

\n
\n

Entrants may re-annotate images in the training set, but may not hand-label predictions, including having human observers rate and evaluate the test data set.

\n
\n

But in this competition I haven't found the same notes about additional(re-labeling) annotations.
\nCan I do bounding bbox annotations of subset of training(only training subset, not test) images or not?
\nIt's not clear for me.
\nThanks.

", "rawMarkdown": "Hello! My question is about of bounding box annotations of training images.\nIn the past [https://www.kaggle.com/c/humpback-whale-identification/ ](url)competition in rules(COMPETITION-SPECIFIC RULES, EXTERNAL DATA section) the following was pointed out:\n> Entrants may re-annotate images in the training set, but may not hand-label predictions, including having human observers rate and evaluate the test data set.\n\nBut in this competition I haven't found the same notes about additional(re-labeling) annotations.\nCan I do bounding bbox annotations of subset of training(only training subset, not test) images or not?\nIt's not clear for me.\nThanks.", "votes": null }, { "id": "1680037", "postDate": "02/07/2022 14:41:06", "content": "

You can annotate the training images. Any test images must be evaluated by your algorithm without any human intervention, so no hand labeling is allowed for the test images.

\n

Here is a discussion with a more detailed answers from kaggle staff

", "rawMarkdown": "You can annotate the training images. Any test images must be evaluated by your algorithm without any human intervention, so no hand labeling is allowed for the test images.\n\nHere is a discussion [with a more detailed answers from kaggle staff](https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/305255)", "votes": null }, { "id": "1685644", "postDate": "02/11/2022 12:51:39", "content": "

There is another question. Are we required to make our new labels publicly available?

", "rawMarkdown": "There is another question. Are we required to make our new labels publicly available?", "votes": null }, { "id": "1686278", "postDate": "02/11/2022 22:32:45", "content": "

This discussion may help you.
\nhttps://www.kaggle.com/c/happy-whale-and-dolphin/discussion/305255#1676925

", "rawMarkdown": "This discussion may help you.\nhttps://www.kaggle.com/c/happy-whale-and-dolphin/discussion/305255#1676925", "votes": null } ], "comments": [ { "id": 1680037, "author_name": "meowmeowmeowmeowmeow", "author_url": "", "post_date": "02/07/2022 14:41:06", "content": "

You can annotate the training images. Any test images must be evaluated by your algorithm without any human intervention, so no hand labeling is allowed for the test images.

\n

Here is a discussion with a more detailed answers from kaggle staff

", "votes": null, "replies": [] }, { "id": 1685644, "author_name": "sggpls", "author_url": "", "post_date": "02/11/2022 12:51:39", "content": "

There is another question. Are we required to make our new labels publicly available?

", "votes": null, "replies": [] }, { "id": 1686278, "author_name": "shinmurashinmura", "author_url": "", "post_date": "02/11/2022 22:32:45", "content": "

This discussion may help you.
\nhttps://www.kaggle.com/c/happy-whale-and-dolphin/discussion/305255#1676925

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "1679838": "Hello! My question is about of bounding box annotations of training images.\nIn the past [https://www.kaggle.com/c/humpback-whale-identification/ ](url)competition in rules(COMPETITION-SPECIFIC RULES, EXTERNAL DATA section) the following was pointed out:\n> Entrants may re-annotate images in the training set, but may not hand-label predictions, including having human observers rate and evaluate the test data set.\n\nBut in this competition I haven't found the same notes about additional(re-labeling) annotations.\nCan I do bounding bbox annotations of subset of training(only training subset, not test) images or not?\nIt's not clear for me.\nThanks.", "1680037": "You can annotate the training images. Any test images must be evaluated by your algorithm without any human intervention, so no hand labeling is allowed for the test images.\n\nHere is a discussion [with a more detailed answers from kaggle staff](https://www.kaggle.com/c/happy-whale-and-dolphin/discussion/305255)", "1685644": "There is another question. Are we required to make our new labels publicly available?", "1686278": "This discussion may help you.\nhttps://www.kaggle.com/c/happy-whale-and-dolphin/discussion/305255#1676925" }, "source": "meta" }