{
  "id": 114294,
  "title": "Softmax vs Sigmoid",
  "url": "/competitions/3d-object-detection-for-autonomous-vehicles/discussion/114294",
  "author_name": "",
  "post_date": "2019-10-25T10:41:34.078291300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>What works better while predicting the masks? Is everyone using <code>Softmax</code> or <code>Sigmoid</code> (class level threshold) for mask prediction?</p>",
  "messages": [
    {
      "id": "657662",
      "postDate": "10/25/2019 10:41:34",
      "content": "<p>What works better while predicting the masks? Is everyone using <code>Softmax</code> or <code>Sigmoid</code> (class level threshold) for mask prediction?</p>",
      "rawMarkdown": "What works better while predicting the masks? Is everyone using `Softmax` or `Sigmoid` (class level threshold) for mask prediction?",
      "votes": null
    },
    {
      "id": "658146",
      "postDate": "10/25/2019 17:54:28",
      "content": "<p>I'm using softmax as each box must belong to a single category. Might I ask what kind of model you're using? Are you classifying the point clouds or the 2D BEV images?</p>",
      "rawMarkdown": "I'm using softmax as each box must belong to a single category. Might I ask what kind of model you're using? Are you classifying the point clouds or the 2D BEV images?",
      "votes": null
    },
    {
      "id": "658872",
      "postDate": "10/26/2019 17:04:36",
      "content": "<p><a href=\"/usmannkhan\">@usmannkhan</a> I started with BEV images, but I am exploring point clouds now</p>",
      "rawMarkdown": "usmannkhan I started with BEV images, but I am exploring point clouds now",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 658146,
      "author_name": "usmannkhan",
      "author_url": "",
      "post_date": "10/25/2019 17:54:28",
      "content": "<p>I'm using softmax as each box must belong to a single category. Might I ask what kind of model you're using? Are you classifying the point clouds or the 2D BEV images?</p>",
      "votes": null,
      "replies": [
        {
          "id": 658872,
          "author_name": "axel81",
          "author_url": "",
          "post_date": "10/26/2019 17:04:36",
          "content": "<p><a href=\"/usmannkhan\">@usmannkhan</a> I started with BEV images, but I am exploring point clouds now</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "657662": "What works better while predicting the masks? Is everyone using `Softmax` or `Sigmoid` (class level threshold) for mask prediction?",
    "658146": "I'm using softmax as each box must belong to a single category. Might I ask what kind of model you're using? Are you classifying the point clouds or the 2D BEV images?",
    "658872": "usmannkhan I started with BEV images, but I am exploring point clouds now"
  },
  "source": "meta"
}