{
  "id": 145884,
  "title": "Provision of evaluation script",
  "url": "/competitions/imaterialist-fashion-2020-fgvc7/discussion/145884",
  "author_name": "Shraddhaa Mohan",
  "post_date": "2020-04-24T23:20:03.424000",
  "votes": 3,
  "comment_count": 6,
  "views": 0,
  "content": "<p>Hey,</p>\n\n<p>Will it be possible to provide the evaluation script used for this competition? Since this is a custom evaluation function, it would be useful to have while conducting experiments on the validation set(created from train)</p>",
  "messages": [
    {
      "id": 819838,
      "postDate": "2020-04-24T23:20:03.423Z",
      "content": "<p>Hey,</p>\n\n<p>Will it be possible to provide the evaluation script used for this competition? Since this is a custom evaluation function, it would be useful to have while conducting experiments on the validation set(created from train)</p>",
      "rawMarkdown": "Hey,\n\nWill it be possible to provide the evaluation script used for this competition? Since this is a custom evaluation function, it would be useful to have while conducting experiments on the validation set(created from train)",
      "votes": 3
    },
    {
      "id": 2230073,
      "postDate": "2023-04-22T01:42:05.193Z",
      "content": "<p>I think it is a cool answer</p>",
      "rawMarkdown": "I think it is a cool answer"
    },
    {
      "id": 932418,
      "postDate": "2020-07-17T03:51:17.177Z",
      "content": "<p>I would recommend you guys using the offical eval metric in <a href=\"https://github.com/KMnP/fashionpedia-api\">Fashionpedia-api.</a>  </p>\n\n<p>It provides a more comprehensive evaluation (using IoU threshold only, and using both IoU and F1 thresholds) so you can understand the performance of your model better.</p>",
      "rawMarkdown": "I would recommend you guys using the offical eval metric in [Fashionpedia-api.](https://github.com/KMnP/fashionpedia-api)  \n\nIt provides a more comprehensive evaluation (using IoU threshold only, and using both IoU and F1 thresholds) so you can understand the performance of your model better."
    },
    {
      "id": 851926,
      "postDate": "2020-05-18T03:26:36.957Z",
      "content": "<p><a href=\"/polosin\">@polosin</a> have you successful write the evaluation script for the competition. I failed it according to the evaluation page. </p>",
      "rawMarkdown": "@polosin have you successful write the evaluation script for the competition. I failed it according to the evaluation page. ",
      "replies": [
        {
          "id": 853078,
          "postDate": "2020-05-19T00:27:58.673Z",
          "content": "<p>I tried to implement the metric the way I understood it, but my cross-validation score didn't match the leaderboard. So at the moment I'm just looking at individual metrics like bbox AP, mask AP and attributes F1-score.</p>",
          "rawMarkdown": "I tried to implement the metric the way I understood it, but my cross-validation score didn't match the leaderboard. So at the moment I'm just looking at individual metrics like bbox AP, mask AP and attributes F1-score."
        }
      ]
    },
    {
      "id": 927187,
      "postDate": "2020-07-13T08:46:25.980Z",
      "rawMarkdown": "",
      "isDeleted": true
    },
    {
      "id": 843858,
      "postDate": "2020-05-12T09:56:01.330Z",
      "rawMarkdown": "",
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 2230073,
      "author_name": "zskitecho",
      "author_url": "",
      "post_date": "2023-04-22T01:42:05.193000",
      "content": "<p>I think it is a cool answer</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 932418,
      "author_name": "Menglin Jia",
      "author_url": "",
      "post_date": "2020-07-17T03:51:17.177000",
      "content": "<p>I would recommend you guys using the offical eval metric in <a href=\"https://github.com/KMnP/fashionpedia-api\">Fashionpedia-api.</a>  </p>\n\n<p>It provides a more comprehensive evaluation (using IoU threshold only, and using both IoU and F1 thresholds) so you can understand the performance of your model better.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 851926,
      "author_name": "Octo",
      "author_url": "",
      "post_date": "2020-05-18T03:26:36.957000",
      "content": "<p><a href=\"/polosin\">@polosin</a> have you successful write the evaluation script for the competition. I failed it according to the evaluation page. </p>",
      "votes": 0,
      "replies": [
        {
          "id": 853078,
          "author_name": "Oleg Polosin",
          "author_url": "",
          "post_date": "2020-05-19T00:27:58.673000",
          "content": "<p>I tried to implement the metric the way I understood it, but my cross-validation score didn't match the leaderboard. So at the moment I'm just looking at individual metrics like bbox AP, mask AP and attributes F1-score.</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 927187,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-13T08:46:25.980000",
      "content": "",
      "votes": 0,
      "replies": []
    },
    {
      "id": 843858,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-05-12T09:56:01.330000",
      "content": "",
      "votes": 0,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "819838": "Hey,\n\nWill it be possible to provide the evaluation script used for this competition? Since this is a custom evaluation function, it would be useful to have while conducting experiments on the validation set(created from train)",
    "2230073": "I think it is a cool answer",
    "932418": "I would recommend you guys using the offical eval metric in [Fashionpedia-api.](https://github.com/KMnP/fashionpedia-api)  \n\nIt provides a more comprehensive evaluation (using IoU threshold only, and using both IoU and F1 thresholds) so you can understand the performance of your model better.",
    "851926": "@polosin have you successful write the evaluation script for the competition. I failed it according to the evaluation page. ",
    "927187": "",
    "843858": ""
  }
}