{
  "id": 64254,
  "title": "Current state of object detection (spoiler: it sucks)",
  "url": "/competitions/google-ai-open-images-object-detection-track/discussion/64254",
  "author_name": "",
  "post_date": "2018-08-27T10:19:53.338090Z",
  "votes": 2,
  "comment_count": 1,
  "views": 0,
  "content": "<p>entertaining work of Rosenfeld et al., check it out the video: <a href=\"https://www.youtube.com/watch?v=qcm3lL4PCC4\">https://www.youtube.com/watch?v=qcm3lL4PCC4</a></p>\n\n<p>the paper: <a href=\"https://arxiv.org/pdf/1808.03305.pdf\">https://arxiv.org/pdf/1808.03305.pdf</a></p>",
  "messages": [
    {
      "id": "376361",
      "postDate": "08/27/2018 10:19:53",
      "content": "<p>entertaining work of Rosenfeld et al., check it out the video: <a href=\"https://www.youtube.com/watch?v=qcm3lL4PCC4\">https://www.youtube.com/watch?v=qcm3lL4PCC4</a></p>\n\n<p>the paper: <a href=\"https://arxiv.org/pdf/1808.03305.pdf\">https://arxiv.org/pdf/1808.03305.pdf</a></p>",
      "rawMarkdown": "entertaining work of Rosenfeld et al., check it out the video: https://www.youtube.com/watch?v=qcm3lL4PCC4\n\nthe paper: https://arxiv.org/pdf/1808.03305.pdf",
      "votes": null
    },
    {
      "id": "376480",
      "postDate": "08/27/2018 15:35:03",
      "content": "<p>If I am not wrong this is the reason Geoffrey Hinton thinks the current CNN architecture, with max-pooling is not a correct representation of how our brain processes images and the fact that it works so well is a disaster. </p>\n\n<p>I hope people come up with innovative CapsNet solutions in the future, they really look amazing on paper and I hope someone gives it a try in this competition( although I think the probability of someone doing it with respectable score is close to none, given the size and complexity of the data set).  </p>",
      "rawMarkdown": "If I am not wrong this is the reason Geoffrey Hinton thinks the current CNN architecture, with max-pooling is not a correct representation of how our brain processes images and the fact that it works so well is a disaster. \n\nI hope people come up with innovative CapsNet solutions in the future, they really look amazing on paper and I hope someone gives it a try in this competition( although I think the probability of someone doing it with respectable score is close to none, given the size and complexity of the data set).",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 376480,
      "author_name": "nazimgirach",
      "author_url": "",
      "post_date": "08/27/2018 15:35:03",
      "content": "<p>If I am not wrong this is the reason Geoffrey Hinton thinks the current CNN architecture, with max-pooling is not a correct representation of how our brain processes images and the fact that it works so well is a disaster. </p>\n\n<p>I hope people come up with innovative CapsNet solutions in the future, they really look amazing on paper and I hope someone gives it a try in this competition( although I think the probability of someone doing it with respectable score is close to none, given the size and complexity of the data set).  </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "376361": "entertaining work of Rosenfeld et al., check it out the video: https://www.youtube.com/watch?v=qcm3lL4PCC4\n\nthe paper: https://arxiv.org/pdf/1808.03305.pdf",
    "376480": "If I am not wrong this is the reason Geoffrey Hinton thinks the current CNN architecture, with max-pooling is not a correct representation of how our brain processes images and the fact that it works so well is a disaster. \n\nI hope people come up with innovative CapsNet solutions in the future, they really look amazing on paper and I hope someone gives it a try in this competition( although I think the probability of someone doing it with respectable score is close to none, given the size and complexity of the data set)."
  },
  "source": "meta"
}