{
  "id": 109537,
  "title": "How does the box score threshold affects the mAP or the final score?",
  "url": "/competitions/open-images-2019-object-detection/discussion/109537",
  "author_name": "",
  "post_date": "2019-09-20T02:36:21.289305300Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Once I tried to revise the box score threshold from 0.0001 to 0.00005, the mAP score demonstrated dramatic decrease. Has anyone realize the issue before? In other words, is it better to submit as many predictions as possible?</p>",
  "messages": [
    {
      "id": "630270",
      "postDate": "09/20/2019 02:36:21",
      "content": "<p>Once I tried to revise the box score threshold from 0.0001 to 0.00005, the mAP score demonstrated dramatic decrease. Has anyone realize the issue before? In other words, is it better to submit as many predictions as possible?</p>",
      "rawMarkdown": "Once I tried to revise the box score threshold from 0.0001 to 0.00005, the mAP score demonstrated dramatic decrease. Has anyone realize the issue before? In other words, is it better to submit as many predictions as possible?",
      "votes": null
    },
    {
      "id": "630271",
      "postDate": "09/20/2019 02:39:17",
      "content": "<p>Does anyone knows the scoring mechanisms?</p>",
      "rawMarkdown": "Does anyone knows the scoring mechanisms?",
      "votes": null
    },
    {
      "id": "630275",
      "postDate": "09/20/2019 02:42:58",
      "content": "<p>It is documented in the challenge and in the dataset. Short version, yes. More predictions are better</p>",
      "rawMarkdown": "It is documented in the challenge and in the dataset. Short version, yes. More predictions are better",
      "votes": null
    },
    {
      "id": "630309",
      "postDate": "09/20/2019 04:22:28",
      "content": "<p>More predictions are better, I believe.</p>",
      "rawMarkdown": "More predictions are better, I believe.",
      "votes": null
    },
    {
      "id": "630327",
      "postDate": "09/20/2019 04:50:17",
      "content": "<p>Thank you. I'll check it!</p>",
      "rawMarkdown": "Thank you. I'll check it!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 630271,
      "author_name": "shixiangyyy",
      "author_url": "",
      "post_date": "09/20/2019 02:39:17",
      "content": "<p>Does anyone knows the scoring mechanisms?</p>",
      "votes": null,
      "replies": [
        {
          "id": 630275,
          "author_name": "moshel",
          "author_url": "",
          "post_date": "09/20/2019 02:42:58",
          "content": "<p>It is documented in the challenge and in the dataset. Short version, yes. More predictions are better</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 630327,
          "author_name": "shixiangyyy",
          "author_url": "",
          "post_date": "09/20/2019 04:50:17",
          "content": "<p>Thank you. I'll check it!</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 630309,
      "author_name": "zorock",
      "author_url": "",
      "post_date": "09/20/2019 04:22:28",
      "content": "<p>More predictions are better, I believe.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "630270": "Once I tried to revise the box score threshold from 0.0001 to 0.00005, the mAP score demonstrated dramatic decrease. Has anyone realize the issue before? In other words, is it better to submit as many predictions as possible?",
    "630271": "Does anyone knows the scoring mechanisms?",
    "630275": "It is documented in the challenge and in the dataset. Short version, yes. More predictions are better",
    "630309": "More predictions are better, I believe.",
    "630327": "Thank you. I'll check it!"
  },
  "source": "meta"
}