![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1528571%2F43771b2c3fd35e39047f55b552604852%2FLB_probing.png?generation=1569149948996763&alt=media\")
Being inspired from this awesome post, I decided to extract similar stats regrading the mask distribution in public test-set by probing the public LB.\nHere is what I got:
\n\n
Useful Stats::
\n\nno. of empty mask = 7055\nno. of non-empty mask = 7737\nno. of non-empty mask for Fish = 1864\nno. of non-empty mask for Flower = 1509\nno. of non-empty mask for Gravel = 1982\nno. of non-empty mask for Sugar = 2382
Above stats were calculated by following Chris comments
\n\nHope this gives you helps you in your experiments and please let me know if my approach/math is wrong\nPS: The numbers are not exact(but close approximation) as I round off decimals.
Being inspired from this awesome post, I decided to extract similar stats regrading the mask distribution in public test-set by probing the public LB.\nHere is what I got:
\n\n
Useful Stats::
\n\nno. of empty mask = 7055\nno. of non-empty mask = 7737\nno. of non-empty mask for Fish = 1864\nno. of non-empty mask for Flower = 1509\nno. of non-empty mask for Gravel = 1982\nno. of non-empty mask for Sugar = 2382
Above stats were calculated by following Chris comments
\n\nHope this gives you helps you in your experiments and please let me know if my approach/math is wrong\nPS: The numbers are not exact(but close approximation) as I round off decimals.
\n\n\nno. of empty mask = 7055\n no. of non-empty mask = 7737\n no. of non-empty mask for
\nFish= 1864\n no. of non-empty mask forFlower= 1509\n no. of non-empty mask forGravel= 1982\n no. of non-empty mask forSugar= 2382
AFAIK the score you got was on 25% of data and these stats assume entire data.
", "rawMarkdown": "> \n> no. of empty mask = 7055\n> no. of non-empty mask = 7737\n> no. of non-empty mask for `Fish` = 1864\n> no. of non-empty mask for `Flower` = 1509\n> no. of non-empty mask for `Gravel` = 1982\n> no. of non-empty mask for `Sugar` = 2382\n\nAFAIK the score you got was on 25% of data and these stats assume entire data.", "votes": null }, { "id": "631676", "postDate": "09/22/2019 13:07:58", "content": "I'm sorry but Could you elaborate on what you mean by 25% of data?
![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F2226462%2F6d6fefaa386f6834beb97703e192cfeb%2FScreenshot%20from%202019-09-22%2020-35-20.png?generation=1569164859746444&alt=media\")
I am talking about this ^
", "rawMarkdown": "\n\nI am talking about this ^", "votes": null }, { "id": "631950", "postDate": "09/23/2019 02:14:13", "content": "![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1528571%2Fc24c3f314c28be29bf95c743f84fcbe0%2Flb.png?generation=1569204717940804&alt=media\")
This is from Steel Competition. The public LB is calculated with 33% of test-data and stats presented by Heng holds for this 33% data.....So I think above stats are valid for 25% of test-data.
", "rawMarkdown": "\n\nThis is from Steel Competition. The public LB is calculated with 33% of test-data and stats presented by Heng holds for this 33% data.....So I think above stats are valid for 25% of test-data.", "votes": null } ], "comments": [ { "id": 631662, "author_name": "dhananjay3", "author_url": "", "post_date": "09/22/2019 12:38:16", "content": "\n\n\nno. of empty mask = 7055\n no. of non-empty mask = 7737\n no. of non-empty mask for
\nFish= 1864\n no. of non-empty mask forFlower= 1509\n no. of non-empty mask forGravel= 1982\n no. of non-empty mask forSugar= 2382
AFAIK the score you got was on 25% of data and these stats assume entire data.
", "votes": null, "replies": [ { "id": 631676, "author_name": "bibek777", "author_url": "", "post_date": "09/22/2019 13:07:58", "content": "I'm sorry but Could you elaborate on what you mean by 25% of data?
I am talking about this ^
", "votes": null, "replies": [] }, { "id": 631950, "author_name": "bibek777", "author_url": "", "post_date": "09/23/2019 02:14:13", "content": "
This is from Steel Competition. The public LB is calculated with 33% of test-data and stats presented by Heng holds for this 33% data.....So I think above stats are valid for 25% of test-data.
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "631634": "Being inspired from this awesome [post](https://www.kaggle.com/c/severstal-steel-defect-detection/discussion/106477#latest-621119), I decided to extract similar stats regrading the mask distribution in public test-set by probing the public LB.\nHere is what I got:\n\n\n\nUseful Stats::\n\nno. of empty mask = 7055\nno. of non-empty mask = 7737\nno. of non-empty mask for `Fish` = 1864\nno. of non-empty mask for `Flower` = 1509\nno. of non-empty mask for `Gravel` = 1982\nno. of non-empty mask for `Sugar` = 2382\n\nAbove stats were calculated by following Chris [comments](https://www.kaggle.com/c/severstal-steel-defect-detection/discussion/106477#619204)\n\nHope this gives you helps you in your experiments and please let me know if my approach/math is wrong\n`PS: The numbers are not exact(but close approximation) as I round off decimals.`", "631662": "> \n> no. of empty mask = 7055\n> no. of non-empty mask = 7737\n> no. of non-empty mask for `Fish` = 1864\n> no. of non-empty mask for `Flower` = 1509\n> no. of non-empty mask for `Gravel` = 1982\n> no. of non-empty mask for `Sugar` = 2382\n\nAFAIK the score you got was on 25% of data and these stats assume entire data.", "631676": "I'm sorry but Could you elaborate on what you mean by `25% of data`?", "631730": "\n\nI am talking about this ^", "631950": "\n\nThis is from Steel Competition. The public LB is calculated with 33% of test-data and stats presented by Heng holds for this 33% data.....So I think above stats are valid for 25% of test-data." }, "source": "meta" }