{
  "id": 156532,
  "title": "Data Augmentation",
  "url": "/competitions/siim-isic-melanoma-classification/discussion/156532",
  "author_name": "",
  "post_date": "2020-06-06T15:28:46.449431300Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I augmented over 10000 images for class 1 and stored it separate folder(under kaggle/working). So now I am using two Image DataGenerators one for class 1 and another for class2</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2F0c6fbe36418822707ae50880bf7c4580%2F0.JPG?generation=1591457226665140&amp;alt=media\" alt=\"\"></p>\n\n<p>So to fit I merged the two generators.\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2Fb0e973e1abe0f33b2ebfea897adb6356%2Fo.JPG?generation=1591457101927242&amp;alt=media\" alt=\"\"></p>\n\n<p>But the problem now is cg generates either 8 images of class 1 or eight images of class 0 as two dataframes are made such that.So the optimizer will be failing to optimize right?</p>\n\n<p>So how to mix the images from both gens, Or should I save the created image augments and given 30000 images of class 1 inside a same directory and continue with a single generator .How to solve my problem?</p>",
  "messages": [
    {
      "id": "876265",
      "postDate": "06/06/2020 15:28:46",
      "content": "<p>I augmented over 10000 images for class 1 and stored it separate folder(under kaggle/working). So now I am using two Image DataGenerators one for class 1 and another for class2</p>\n\n<p><img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2F0c6fbe36418822707ae50880bf7c4580%2F0.JPG?generation=1591457226665140&amp;alt=media\" alt=\"\"></p>\n\n<p>So to fit I merged the two generators.\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2Fb0e973e1abe0f33b2ebfea897adb6356%2Fo.JPG?generation=1591457101927242&amp;alt=media\" alt=\"\"></p>\n\n<p>But the problem now is cg generates either 8 images of class 1 or eight images of class 0 as two dataframes are made such that.So the optimizer will be failing to optimize right?</p>\n\n<p>So how to mix the images from both gens, Or should I save the created image augments and given 30000 images of class 1 inside a same directory and continue with a single generator .How to solve my problem?</p>",
      "rawMarkdown": "I augmented over 10000 images for class 1 and stored it separate folder(under kaggle/working). So now I am using two Image DataGenerators one for class 1 and another for class2\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2F0c6fbe36418822707ae50880bf7c4580%2F0.JPG?generation=1591457226665140&amp;alt=media)\n\n\n\n So to fit I merged the two generators.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2Fb0e973e1abe0f33b2ebfea897adb6356%2Fo.JPG?generation=1591457101927242&amp;alt=media)\n\n\nBut the problem now is cg generates either 8 images of class 1 or eight images of class 0 as two dataframes are made such that.So the optimizer will be failing to optimize right?\n\nSo how to mix the images from both gens, Or should I save the created image augments and given 30000 images of class 1 inside a same directory and continue with a single generator .How to solve my problem?",
      "votes": null
    },
    {
      "id": "888298",
      "postDate": "06/16/2020 09:04:39",
      "content": "<p>Same here. I want to store images using data augmentation but don't know how to store those images.</p>",
      "rawMarkdown": "Same here. I want to store images using data augmentation but don't know how to store those images.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 888298,
      "author_name": "sanketpatel108784",
      "author_url": "",
      "post_date": "06/16/2020 09:04:39",
      "content": "<p>Same here. I want to store images using data augmentation but don't know how to store those images.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "876265": "I augmented over 10000 images for class 1 and stored it separate folder(under kaggle/working). So now I am using two Image DataGenerators one for class 1 and another for class2\n\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2F0c6fbe36418822707ae50880bf7c4580%2F0.JPG?generation=1591457226665140&amp;alt=media)\n\n\n\n So to fit I merged the two generators.\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3597887%2Fb0e973e1abe0f33b2ebfea897adb6356%2Fo.JPG?generation=1591457101927242&amp;alt=media)\n\n\nBut the problem now is cg generates either 8 images of class 1 or eight images of class 0 as two dataframes are made such that.So the optimizer will be failing to optimize right?\n\nSo how to mix the images from both gens, Or should I save the created image augments and given 30000 images of class 1 inside a same directory and continue with a single generator .How to solve my problem?",
    "888298": "Same here. I want to store images using data augmentation but don't know how to store those images."
  },
  "source": "meta"
}