{
  "id": 100355,
  "title": "Private LB Simulation [5 min read]",
  "url": "/competitions/generative-dog-images/discussion/100355",
  "author_name": "",
  "post_date": "2019-07-18T02:21:49.817045700Z",
  "votes": 6,
  "comment_count": 2,
  "views": 0,
  "content": "<h1><a href=\"https://www.kaggle.com/jesucristo/private-lb-simulation-i\">Private LB Simulation I</a></h1>\n\n<p></p>\n\n<p>I'm back! no more discussions about memorizers or rules, it's boring.</p>\n\n<p><strong>I think this can help everyone, no matter what kind of GAN you use: memorizer, GAN, CGAN, RaLSGAN etc</strong></p>\n\n<p><strong>In <a href=\"https://www.kaggle.com/jesucristo/private-lb-simulation-i\">my kernel</a> I simulate the private dataset!</strong> And you'll see if the dataset we use to generate images matters when it comes to getting a good LB.</p>\n\n<h3>The question is ...</h3>\n\n<p>(or at least I asked myself this 😄 )</p>\n\n<ul>\n<li><code>private dataset = all dogs</code>, generate using <code>X</code> gives us a LB around 7.5 - 8.5</li>\n<li><strong><code>private dataset = X</code>, generate using <code>all dogs</code> gives us a LB around 7.5 - 8.5 ? This is the real case, and I think the answer is yes!</strong></li>\n</ul>\n\n<p>so <strong>what is X?</strong></p>\n\n<p>In <a href=\"https://www.kaggle.com/jesucristo/private-lb-simulation-i\">my kernel</a> you'll see a dataset with the following description:</p>\n\n<blockquote>\n  <p>... strictly canine subset of ImageNet in order to practice fine-grained image categorization. How well you can tell your Norfolk Terriers from your Norwich Terriers? With 120 breeds of dogs ... the creators of the Stanford Dogs Dataset for making this possible: Aditya Khosla, Nityananda Jayadevaprakash, Bangpeng Yao, and Fei-Fei Li.</p>\n</blockquote>\n\n<p>And contains images like this:</p>\n\n<h2><img src=\"https://storage.googleapis.com/kaggle-competitions/kaggle/3333/media/border_collies.png\" alt=\"\"></h2>\n\n<blockquote>\n  <p>This experiment/simulation took me some time, so I hope you'll appreciate it :)</p>\n</blockquote>",
  "messages": [
    {
      "id": "578667",
      "postDate": "07/18/2019 02:21:49",
      "content": "<h1><a href=\"https://www.kaggle.com/jesucristo/private-lb-simulation-i\">Private LB Simulation I</a></h1>\n\n<p></p>\n\n<p>I'm back! no more discussions about memorizers or rules, it's boring.</p>\n\n<p><strong>I think this can help everyone, no matter what kind of GAN you use: memorizer, GAN, CGAN, RaLSGAN etc</strong></p>\n\n<p><strong>In <a href=\"https://www.kaggle.com/jesucristo/private-lb-simulation-i\">my kernel</a> I simulate the private dataset!</strong> And you'll see if the dataset we use to generate images matters when it comes to getting a good LB.</p>\n\n<h3>The question is ...</h3>\n\n<p>(or at least I asked myself this 😄 )</p>\n\n<ul>\n<li><code>private dataset = all dogs</code>, generate using <code>X</code> gives us a LB around 7.5 - 8.5</li>\n<li><strong><code>private dataset = X</code>, generate using <code>all dogs</code> gives us a LB around 7.5 - 8.5 ? This is the real case, and I think the answer is yes!</strong></li>\n</ul>\n\n<p>so <strong>what is X?</strong></p>\n\n<p>In <a href=\"https://www.kaggle.com/jesucristo/private-lb-simulation-i\">my kernel</a> you'll see a dataset with the following description:</p>\n\n<blockquote>\n  <p>... strictly canine subset of ImageNet in order to practice fine-grained image categorization. How well you can tell your Norfolk Terriers from your Norwich Terriers? With 120 breeds of dogs ... the creators of the Stanford Dogs Dataset for making this possible: Aditya Khosla, Nityananda Jayadevaprakash, Bangpeng Yao, and Fei-Fei Li.</p>\n</blockquote>\n\n<p>And contains images like this:</p>\n\n<h2><img src=\"https://storage.googleapis.com/kaggle-competitions/kaggle/3333/media/border_collies.png\" alt=\"\"></h2>\n\n<blockquote>\n  <p>This experiment/simulation took me some time, so I hope you'll appreciate it :)</p>\n</blockquote>",
      "rawMarkdown": "# [Private LB Simulation I](https://www.kaggle.com/jesucristo/private-lb-simulation-i)\n\n<img src=\"https://i.ibb.co/DVn018V/Captura-de-pantalla-de-2019-07-18-02-55-09.png\" alt=\"Captura-de-pantalla-de-2019-07-18-02-55-09\">\n\n\nI'm back! no more discussions about memorizers or rules, it's boring.\n\n**I think this can help everyone, no matter what kind of GAN you use: memorizer, GAN, CGAN, RaLSGAN etc**\n\n**In [my kernel](https://www.kaggle.com/jesucristo/private-lb-simulation-i) I simulate the private dataset!** And you'll see if the dataset we use to generate images matters when it comes to getting a good LB.\n\n### The question is ...\n\n(or at least I asked myself this 😄 )\n\n- ```private dataset = all dogs```, generate using ```X``` gives us a LB around 7.5 - 8.5\n- **```private dataset = X```, generate using ```all dogs``` gives us a LB around 7.5 - 8.5 ? This is the real case, and I think the answer is yes!**\n\nso **what is X?**\n\nIn [my kernel](https://www.kaggle.com/jesucristo/private-lb-simulation-i) you'll see a dataset with the following description:\n\n&gt; ... strictly canine subset of ImageNet in order to practice fine-grained image categorization. How well you can tell your Norfolk Terriers from your Norwich Terriers? With 120 breeds of dogs ... the creators of the Stanford Dogs Dataset for making this possible: Aditya Khosla, Nityananda Jayadevaprakash, Bangpeng Yao, and Fei-Fei Li.\n\nAnd contains images like this:\n\n![](https://storage.googleapis.com/kaggle-competitions/kaggle/3333/media/border_collies.png)\n---\n\n&gt; This experiment/simulation took me some time, so I hope you'll appreciate it :)",
      "votes": null
    },
    {
      "id": "578674",
      "postDate": "07/18/2019 02:32:43",
      "content": "<p>Thanks <a href=\"/jesucristo\">@jesucristo</a> for sharing ...this is useful.</p>",
      "rawMarkdown": "Thanks @jesucristo for sharing ...this is useful.",
      "votes": null
    },
    {
      "id": "578677",
      "postDate": "07/18/2019 02:37:33",
      "content": "<p>👍 glad that you liked it ;)</p>",
      "rawMarkdown": "👍 glad that you liked it ;)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 578674,
      "author_name": "prmohanty",
      "author_url": "",
      "post_date": "07/18/2019 02:32:43",
      "content": "<p>Thanks <a href=\"/jesucristo\">@jesucristo</a> for sharing ...this is useful.</p>",
      "votes": null,
      "replies": [
        {
          "id": 578677,
          "author_name": "jesucristo",
          "author_url": "",
          "post_date": "07/18/2019 02:37:33",
          "content": "<p>👍 glad that you liked it ;)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "578667": "# [Private LB Simulation I](https://www.kaggle.com/jesucristo/private-lb-simulation-i)\n\n<img src=\"https://i.ibb.co/DVn018V/Captura-de-pantalla-de-2019-07-18-02-55-09.png\" alt=\"Captura-de-pantalla-de-2019-07-18-02-55-09\">\n\n\nI'm back! no more discussions about memorizers or rules, it's boring.\n\n**I think this can help everyone, no matter what kind of GAN you use: memorizer, GAN, CGAN, RaLSGAN etc**\n\n**In [my kernel](https://www.kaggle.com/jesucristo/private-lb-simulation-i) I simulate the private dataset!** And you'll see if the dataset we use to generate images matters when it comes to getting a good LB.\n\n### The question is ...\n\n(or at least I asked myself this 😄 )\n\n- ```private dataset = all dogs```, generate using ```X``` gives us a LB around 7.5 - 8.5\n- **```private dataset = X```, generate using ```all dogs``` gives us a LB around 7.5 - 8.5 ? This is the real case, and I think the answer is yes!**\n\nso **what is X?**\n\nIn [my kernel](https://www.kaggle.com/jesucristo/private-lb-simulation-i) you'll see a dataset with the following description:\n\n&gt; ... strictly canine subset of ImageNet in order to practice fine-grained image categorization. How well you can tell your Norfolk Terriers from your Norwich Terriers? With 120 breeds of dogs ... the creators of the Stanford Dogs Dataset for making this possible: Aditya Khosla, Nityananda Jayadevaprakash, Bangpeng Yao, and Fei-Fei Li.\n\nAnd contains images like this:\n\n![](https://storage.googleapis.com/kaggle-competitions/kaggle/3333/media/border_collies.png)\n---\n\n&gt; This experiment/simulation took me some time, so I hope you'll appreciate it :)",
    "578674": "Thanks @jesucristo for sharing ...this is useful.",
    "578677": "👍 glad that you liked it ;)"
  },
  "source": "meta"
}