{ "id": 119632, "title": "Why are the scores in the top of the rankings only one significant number？", "url": "/competitions/pku-autonomous-driving/discussion/119632", "author_name": "", "post_date": "2019-11-30T03:34:14.353086100Z", "votes": 4, "comment_count": 11, "views": 0, "content": "

\"\"

", "messages": [ { "id": "684590", "postDate": "11/30/2019 03:34:14", "content": "

\"\"

", "rawMarkdown": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4060903%2F837c95ca76091e435bdd151f2535ff28%2FScreenshot%20from%202019-11-30%2011-33-37.png?generation=1575084849473767&alt=media)", "votes": null }, { "id": "684623", "postDate": "11/30/2019 04:52:13", "content": "

This is quite strange, very low chance this could be correct.

", "rawMarkdown": "This is quite strange, very low chance this could be correct.", "votes": null }, { "id": "684633", "postDate": "11/30/2019 05:34:26", "content": "

It's just so much better than the others ...that Kaggle dont even bother to calculate 3 decimals :P

", "rawMarkdown": "It's just so much better than the others ...that Kaggle dont even bother to calculate 3 decimals :P", "votes": null }, { "id": "684757", "postDate": "11/30/2019 12:19:42", "content": "

The speed of problem solving by the organizers is also strange

", "rawMarkdown": "The speed of problem solving by the organizers is also strange", "votes": null }, { "id": "684776", "postDate": "11/30/2019 13:04:05", "content": "
    \n
  1. Scores are only calculated on 181 images 2. Metric to calculate score is wrong and doesn't take into account unlabeled cars. All those scores are from predictions that label around 10 cars or less.
    2. \n
", "rawMarkdown": "1. Scores are only calculated on 181 images 2. Metric to calculate score is wrong and doesn't take into account unlabeled cars. All those scores are from predictions that label around 10 cars or less.", "votes": null }, { "id": "684797", "postDate": "11/30/2019 13:48:48", "content": "

The score was supposed to change this week right ? I was waiting for it eagerly . Just to get an idea how good or how bad are my models , with respect to competition goal

", "rawMarkdown": "The score was supposed to change this week right ? I was waiting for it eagerly . Just to get an idea how good or how bad are my models , with respect to competition goal", "votes": null }, { "id": "684865", "postDate": "11/30/2019 16:21:13", "content": "

@greatgamedota So to get a perfect score on public LB you just need to get one correct object at each threshold?

", "rawMarkdown": "greatgamedota So to get a perfect score on public LB you just need to get one correct object at each threshold?", "votes": null }, { "id": "684867", "postDate": "11/30/2019 16:22:42", "content": "

Very echo friendly of them πŸ˜‚

", "rawMarkdown": "Very echo friendly of them πŸ˜‚", "votes": null }, { "id": "684955", "postDate": "11/30/2019 17:30:55", "content": "

πŸ˜„

", "rawMarkdown": "πŸ˜„", "votes": null }, { "id": "684993", "postDate": "11/30/2019 18:53:28", "content": "

No 2 and 3 also look somewhat like a team to me...

\n\n

Please note, private sharing outside of teams is forbidden.

", "rawMarkdown": "No 2 and 3 also look somewhat like a team to me...\n\nPlease note, private sharing outside of teams is forbidden.", "votes": null }, { "id": "684994", "postDate": "11/30/2019 18:57:21", "content": "

Check vs the public scoring kernel, @phoenix9032 \nAnything above 0.150 is pretty good I guess

", "rawMarkdown": "Check vs the public scoring kernel, @phoenix9032 \nAnything above 0.150 is pretty good I guess", "votes": null }, { "id": "685002", "postDate": "11/30/2019 19:22:06", "content": "

Thanks @ilu000 , that is more inline with the kitti benchmarks.

", "rawMarkdown": "Thanks @ilu000 , that is more inline with the kitti benchmarks.", "votes": null } ], "comments": [ { "id": 684623, "author_name": "jackvial", "author_url": "", "post_date": "11/30/2019 04:52:13", "content": "

This is quite strange, very low chance this could be correct.

", "votes": null, "replies": [ { "id": 684633, "author_name": "phoenix9032", "author_url": "", "post_date": "11/30/2019 05:34:26", "content": "

It's just so much better than the others ...that Kaggle dont even bother to calculate 3 decimals :P

", "votes": null, "replies": [] }, { "id": 684867, "author_name": "jackvial", "author_url": "", "post_date": "11/30/2019 16:22:42", "content": "

Very echo friendly of them πŸ˜‚

", "votes": null, "replies": [] }, { "id": 684955, "author_name": "phoenix9032", "author_url": "", "post_date": "11/30/2019 17:30:55", "content": "

πŸ˜„

", "votes": null, "replies": [] } ] }, { "id": 684757, "author_name": "frkaka", "author_url": "", "post_date": "11/30/2019 12:19:42", "content": "

The speed of problem solving by the organizers is also strange

", "votes": null, "replies": [] }, { "id": 684776, "author_name": "greatgamedota", "author_url": "", "post_date": "11/30/2019 13:04:05", "content": "
    \n
  1. Scores are only calculated on 181 images 2. Metric to calculate score is wrong and doesn't take into account unlabeled cars. All those scores are from predictions that label around 10 cars or less.
    2. \n
", "votes": null, "replies": [ { "id": 684797, "author_name": "phoenix9032", "author_url": "", "post_date": "11/30/2019 13:48:48", "content": "

The score was supposed to change this week right ? I was waiting for it eagerly . Just to get an idea how good or how bad are my models , with respect to competition goal

", "votes": null, "replies": [] }, { "id": 684865, "author_name": "jackvial", "author_url": "", "post_date": "11/30/2019 16:21:13", "content": "

@greatgamedota So to get a perfect score on public LB you just need to get one correct object at each threshold?

", "votes": null, "replies": [] }, { "id": 684994, "author_name": "ilu000", "author_url": "", "post_date": "11/30/2019 18:57:21", "content": "

Check vs the public scoring kernel, @phoenix9032 \nAnything above 0.150 is pretty good I guess

", "votes": null, "replies": [] }, { "id": 685002, "author_name": "jackvial", "author_url": "", "post_date": "11/30/2019 19:22:06", "content": "

Thanks @ilu000 , that is more inline with the kitti benchmarks.

", "votes": null, "replies": [] } ] }, { "id": 684993, "author_name": "ilu000", "author_url": "", "post_date": "11/30/2019 18:53:28", "content": "

No 2 and 3 also look somewhat like a team to me...

\n\n

Please note, private sharing outside of teams is forbidden.

", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "684590": "![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F4060903%2F837c95ca76091e435bdd151f2535ff28%2FScreenshot%20from%202019-11-30%2011-33-37.png?generation=1575084849473767&alt=media)", "684623": "This is quite strange, very low chance this could be correct.", "684633": "It's just so much better than the others ...that Kaggle dont even bother to calculate 3 decimals :P", "684757": "The speed of problem solving by the organizers is also strange", "684776": "1. Scores are only calculated on 181 images 2. Metric to calculate score is wrong and doesn't take into account unlabeled cars. All those scores are from predictions that label around 10 cars or less.", "684797": "The score was supposed to change this week right ? I was waiting for it eagerly . Just to get an idea how good or how bad are my models , with respect to competition goal", "684865": "greatgamedota So to get a perfect score on public LB you just need to get one correct object at each threshold?", "684867": "Very echo friendly of them πŸ˜‚", "684955": "πŸ˜„", "684993": "No 2 and 3 also look somewhat like a team to me...\n\nPlease note, private sharing outside of teams is forbidden.", "684994": "Check vs the public scoring kernel, @phoenix9032 \nAnything above 0.150 is pretty good I guess", "685002": "Thanks @ilu000 , that is more inline with the kitti benchmarks." }, "source": "meta" }