{
  "id": 128139,
  "title": "Generic vs Specific model",
  "url": "/competitions/bengaliai-cv19/discussion/128139",
  "author_name": "Son of Anton v3.0",
  "post_date": "2020-01-29T07:19:44.149000",
  "votes": 1,
  "comment_count": 0,
  "views": 0,
  "content": "<p>The competition requires 3 separate classification tasks to be done. Intuitively, it would make sense to separate models for different tasks as different features might be extracted. But many have used a single model to extract the features with a multi-output CNN and it has paid rich dividends. So which approach is suitable for this problem? In the case of a generic model, what is happening behind the hood?</p>",
  "messages": [
    {
      "id": 731881,
      "postDate": "2020-01-29T07:19:44.150Z",
      "content": "<p>The competition requires 3 separate classification tasks to be done. Intuitively, it would make sense to separate models for different tasks as different features might be extracted. But many have used a single model to extract the features with a multi-output CNN and it has paid rich dividends. So which approach is suitable for this problem? In the case of a generic model, what is happening behind the hood?</p>",
      "rawMarkdown": "The competition requires 3 separate classification tasks to be done. Intuitively, it would make sense to separate models for different tasks as different features might be extracted. But many have used a single model to extract the features with a multi-output CNN and it has paid rich dividends. So which approach is suitable for this problem? In the case of a generic model, what is happening behind the hood?",
      "votes": 1
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "731881": "The competition requires 3 separate classification tasks to be done. Intuitively, it would make sense to separate models for different tasks as different features might be extracted. But many have used a single model to extract the features with a multi-output CNN and it has paid rich dividends. So which approach is suitable for this problem? In the case of a generic model, what is happening behind the hood?"
  }
}