{ "id": 105671, "title": "Mistake in `challenge-2019-label300-segmentable-hierarchy.json`?", "url": "/competitions/open-images-2019-instance-segmentation/discussion/105671", "author_name": "", "post_date": "2019-08-25T09:55:15.113953100Z", "votes": 2, "comment_count": 5, "views": 0, "content": "

In the file challenge-2019-label300-segmentable-hierarchy.json which I got from here, the very first label is given as /m/0bl9f, which doesn't seem to exist in challenge-2019-classes-description-segmentable.csv file which you can find here

\n\n

Another thing to note is that every other category is part of this label, i.e /m/0bl9f. \nAny clarification on the same would be really helpful!

", "messages": [ { "id": "607463", "postDate": "08/25/2019 09:55:15", "content": "

In the file challenge-2019-label300-segmentable-hierarchy.json which I got from here, the very first label is given as /m/0bl9f, which doesn't seem to exist in challenge-2019-classes-description-segmentable.csv file which you can find here

\n\n

Another thing to note is that every other category is part of this label, i.e /m/0bl9f. \nAny clarification on the same would be really helpful!

", "rawMarkdown": "In the file `challenge-2019-label300-segmentable-hierarchy.json` which I got from [here](https://storage.googleapis.com/openimages/challenge_2019/challenge-2019-label300-segmentable-hierarchy.json), the very first label is given as `/m/0bl9f`, which doesn't seem to exist in `challenge-2019-classes-description-segmentable.csv` file which you can find [here](https://storage.googleapis.com/openimages/challenge_2019/challenge-2019-classes-description-segmentable.csv)\n\nAnother thing to note is that every other category is part of this label, i.e `/m/0bl9f`. \nAny clarification on the same would be really helpful!", "votes": null }, { "id": "609065", "postDate": "08/27/2019 11:45:23", "content": "

From the semantic hierarchy visualisation, I take it that class is just a hierarchy root which Google named Entity.\nFor the purpose of the challenge, as one can read at the challenge evaluation page, in the section 'Labels hierarchy handling',

\n\n
\n

Note: the root 'Entity' class is not part of the challenge and thus is not evaluated.

\n
\n\n

The dataset information is somewhat nontrivial to retrieve... But I guess you figured all of this out in the meantime? =)

", "rawMarkdown": "From the [semantic hierarchy visualisation](https://storage.googleapis.com/openimages/challenge_2018/bbox_labels_500_hierarchy_visualizer/circle.html), I take it that class is just a hierarchy root which Google named `Entity`.\nFor the purpose of the challenge, as one can read at the [challenge evaluation page](https://storage.googleapis.com/openimages/web/evaluation.html), in the section 'Labels hierarchy handling',\n> Note: the root 'Entity' class is not part of the challenge and thus is not evaluated.\n\nThe dataset information is somewhat nontrivial to retrieve... But I guess you figured all of this out in the meantime? =)", "votes": null }, { "id": "610977", "postDate": "08/29/2019 03:42:38", "content": "

Yup, I missed out on that Note. \nThanks for your reply but 😄

", "rawMarkdown": "Yup, I missed out on that Note. \nThanks for your reply but 😄", "votes": null }, { "id": "615127", "postDate": "09/01/2019 13:45:32", "content": "

Hey @lntsmn if you dont mind clarifying another thing for me, the Person class seems to be a super category but has lesser counts than the Man class. Any idea how to deal with such an oddity?

", "rawMarkdown": "Hey @lntsmn if you dont mind clarifying another thing for me, the `Person` class seems to be a super category but has lesser counts than the `Man` class. Any idea how to deal with such an oddity?", "votes": null }, { "id": "615993", "postDate": "09/02/2019 15:12:47", "content": "

Hi @rohitmidha23, sorry for the delay...\nAs far as I understand, objects are annotated with their deepest known subclasses, i.e. if a Person is recognised as a Man only Man will be annotated, if uncertain about age/gender it will just be marked as Person. It would seem that the cases where a definite Man appeared in an image and was recognised as such outnumbered the cases where age or sex was uncertain.

", "rawMarkdown": "Hi @rohitmidha23, sorry for the delay...\nAs far as I understand, objects are annotated with their deepest known subclasses, i.e. if a `Person` is recognised as a `Man` only `Man` will be annotated, if uncertain about age/gender it will just be marked as `Person`. It would seem that the cases where a definite `Man` appeared in an image and was recognised as such outnumbered the cases where age or sex was uncertain.", "votes": null }, { "id": "616534", "postDate": "09/03/2019 08:06:09", "content": "

I see, thanks so much.

", "rawMarkdown": "I see, thanks so much.", "votes": null } ], "comments": [ { "id": 609065, "author_name": "lntsmn", "author_url": "", "post_date": "08/27/2019 11:45:23", "content": "

From the semantic hierarchy visualisation, I take it that class is just a hierarchy root which Google named Entity.\nFor the purpose of the challenge, as one can read at the challenge evaluation page, in the section 'Labels hierarchy handling',

\n\n
\n

Note: the root 'Entity' class is not part of the challenge and thus is not evaluated.

\n
\n\n

The dataset information is somewhat nontrivial to retrieve... But I guess you figured all of this out in the meantime? =)

", "votes": null, "replies": [ { "id": 610977, "author_name": "rohitmidha23", "author_url": "", "post_date": "08/29/2019 03:42:38", "content": "

Yup, I missed out on that Note. \nThanks for your reply but 😄

", "votes": null, "replies": [] }, { "id": 615127, "author_name": "rohitmidha23", "author_url": "", "post_date": "09/01/2019 13:45:32", "content": "

Hey @lntsmn if you dont mind clarifying another thing for me, the Person class seems to be a super category but has lesser counts than the Man class. Any idea how to deal with such an oddity?

", "votes": null, "replies": [] }, { "id": 615993, "author_name": "lntsmn", "author_url": "", "post_date": "09/02/2019 15:12:47", "content": "

Hi @rohitmidha23, sorry for the delay...\nAs far as I understand, objects are annotated with their deepest known subclasses, i.e. if a Person is recognised as a Man only Man will be annotated, if uncertain about age/gender it will just be marked as Person. It would seem that the cases where a definite Man appeared in an image and was recognised as such outnumbered the cases where age or sex was uncertain.

", "votes": null, "replies": [] }, { "id": 616534, "author_name": "rohitmidha23", "author_url": "", "post_date": "09/03/2019 08:06:09", "content": "

I see, thanks so much.

", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "607463": "In the file `challenge-2019-label300-segmentable-hierarchy.json` which I got from [here](https://storage.googleapis.com/openimages/challenge_2019/challenge-2019-label300-segmentable-hierarchy.json), the very first label is given as `/m/0bl9f`, which doesn't seem to exist in `challenge-2019-classes-description-segmentable.csv` file which you can find [here](https://storage.googleapis.com/openimages/challenge_2019/challenge-2019-classes-description-segmentable.csv)\n\nAnother thing to note is that every other category is part of this label, i.e `/m/0bl9f`. \nAny clarification on the same would be really helpful!", "609065": "From the [semantic hierarchy visualisation](https://storage.googleapis.com/openimages/challenge_2018/bbox_labels_500_hierarchy_visualizer/circle.html), I take it that class is just a hierarchy root which Google named `Entity`.\nFor the purpose of the challenge, as one can read at the [challenge evaluation page](https://storage.googleapis.com/openimages/web/evaluation.html), in the section 'Labels hierarchy handling',\n> Note: the root 'Entity' class is not part of the challenge and thus is not evaluated.\n\nThe dataset information is somewhat nontrivial to retrieve... But I guess you figured all of this out in the meantime? =)", "610977": "Yup, I missed out on that Note. \nThanks for your reply but 😄", "615127": "Hey @lntsmn if you dont mind clarifying another thing for me, the `Person` class seems to be a super category but has lesser counts than the `Man` class. Any idea how to deal with such an oddity?", "615993": "Hi @rohitmidha23, sorry for the delay...\nAs far as I understand, objects are annotated with their deepest known subclasses, i.e. if a `Person` is recognised as a `Man` only `Man` will be annotated, if uncertain about age/gender it will just be marked as `Person`. It would seem that the cases where a definite `Man` appeared in an image and was recognised as such outnumbered the cases where age or sex was uncertain.", "616534": "I see, thanks so much." }, "source": "meta" }