{
  "id": 98488,
  "title": "How to measure my GAN is performing well ?",
  "url": "/competitions/generative-dog-images/discussion/98488",
  "author_name": "",
  "post_date": "2019-07-04T05:05:55.263409Z",
  "votes": 1,
  "comment_count": 3,
  "views": 0,
  "content": "<p>Can anyone tell me how do I measure If my GAN is performing well than previously trained GAN.</p>\n\n<p>The generated images are not as good as I can see and tell it's working better. </p>\n\n<p>Submitting to competition is an option but that takes a lot of time.</p>\n\n<p>I tried plotting loss functions but that doesn't work as well.</p>",
  "messages": [
    {
      "id": "567861",
      "postDate": "07/04/2019 05:05:55",
      "content": "<p>Can anyone tell me how do I measure If my GAN is performing well than previously trained GAN.</p>\n\n<p>The generated images are not as good as I can see and tell it's working better. </p>\n\n<p>Submitting to competition is an option but that takes a lot of time.</p>\n\n<p>I tried plotting loss functions but that doesn't work as well.</p>",
      "rawMarkdown": "Can anyone tell me how do I measure If my GAN is performing well than previously trained GAN.\n\nThe generated images are not as good as I can see and tell it's working better. \n\nSubmitting to competition is an option but that takes a lot of time.\n\nI tried plotting loss functions but that doesn't work as well.",
      "votes": null
    },
    {
      "id": "568123",
      "postDate": "07/04/2019 12:43:05",
      "content": "<p>Easy, use the FID metric,see this kernel <a href=\"https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp\">https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp</a>, use your generated images there and that's all :) </p>",
      "rawMarkdown": "Easy, use the FID metric,see this kernel https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp, use your generated images there and that's all :)",
      "votes": null
    },
    {
      "id": "568129",
      "postDate": "07/04/2019 12:56:17",
      "content": "<p>There is not a single metric for measuring the performance of GANs while testing although FID does seems promising. Although it does not help with the problem of <em>Mode Collapse</em> and <em>Moemorization</em> which are frequent in GAN.</p>",
      "rawMarkdown": "There is not a single metric for measuring the performance of GANs while testing although FID does seems promising. Although it does not help with the problem of _Mode Collapse_ and _Moemorization_ which are frequent in GAN.",
      "votes": null
    },
    {
      "id": "568521",
      "postDate": "07/05/2019 03:56:37",
      "content": "<p>ok thanks !</p>",
      "rawMarkdown": "ok thanks !",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 568123,
      "author_name": "jesucristo",
      "author_url": "",
      "post_date": "07/04/2019 12:43:05",
      "content": "<p>Easy, use the FID metric,see this kernel <a href=\"https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp\">https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp</a>, use your generated images there and that's all :) </p>",
      "votes": null,
      "replies": [
        {
          "id": 568521,
          "author_name": "whizzkid",
          "author_url": "",
          "post_date": "07/05/2019 03:56:37",
          "content": "<p>ok thanks !</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 568129,
      "author_name": "thanatoz",
      "author_url": "",
      "post_date": "07/04/2019 12:56:17",
      "content": "<p>There is not a single metric for measuring the performance of GANs while testing although FID does seems promising. Although it does not help with the problem of <em>Mode Collapse</em> and <em>Moemorization</em> which are frequent in GAN.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "567861": "Can anyone tell me how do I measure If my GAN is performing well than previously trained GAN.\n\nThe generated images are not as good as I can see and tell it's working better. \n\nSubmitting to competition is an option but that takes a lot of time.\n\nI tried plotting loss functions but that doesn't work as well.",
    "568123": "Easy, use the FID metric,see this kernel https://www.kaggle.com/wendykan/demo-mifid-metric-for-dog-image-generation-comp, use your generated images there and that's all :)",
    "568129": "There is not a single metric for measuring the performance of GANs while testing although FID does seems promising. Although it does not help with the problem of _Mode Collapse_ and _Moemorization_ which are frequent in GAN.",
    "568521": "ok thanks !"
  },
  "source": "meta"
}