{
  "id": 163553,
  "title": "Can You get a good score without using all images ? ",
  "url": "/competitions/alaska2-image-steganalysis/discussion/163553",
  "author_name": "Mohamed Annis SOUAMES",
  "post_date": "2020-07-02T13:25:18.535000",
  "votes": 0,
  "comment_count": 1,
  "views": 0,
  "content": "<p>This is my first Kaggle competition and my first one using computer vision (I only worked on tabular data previously) and I'm wondering whether it's possible to get a better score than 0.92 using only a given fraction of the images ? </p>\n\n<p>I'm asking this because the data is very large and it's infeasible to train CNN like EffNet B(0-7) and others using all images and Kaggle runtime. I can't use my computer since I don't have a powerful GPU just 4 GB of RAM and I5 for CPU and my internet connection is very slow (2MB/s), so the only way for me to learn and participate in such competitions is using Kaggle Runtime or Colab.  </p>",
  "messages": [
    {
      "id": 912437,
      "postDate": "2020-07-02T13:25:18.537Z",
      "content": "<p>This is my first Kaggle competition and my first one using computer vision (I only worked on tabular data previously) and I'm wondering whether it's possible to get a better score than 0.92 using only a given fraction of the images ? </p>\n\n<p>I'm asking this because the data is very large and it's infeasible to train CNN like EffNet B(0-7) and others using all images and Kaggle runtime. I can't use my computer since I don't have a powerful GPU just 4 GB of RAM and I5 for CPU and my internet connection is very slow (2MB/s), so the only way for me to learn and participate in such competitions is using Kaggle Runtime or Colab.  </p>",
      "rawMarkdown": "This is my first Kaggle competition and my first one using computer vision (I only worked on tabular data previously) and I'm wondering whether it's possible to get a better score than 0.92 using only a given fraction of the images ? \n\nI'm asking this because the data is very large and it's infeasible to train CNN like EffNet B(0-7) and others using all images and Kaggle runtime. I can't use my computer since I don't have a powerful GPU just 4 GB of RAM and I5 for CPU and my internet connection is very slow (2MB/s), so the only way for me to learn and participate in such competitions is using Kaggle Runtime or Colab.  "
    },
    {
      "id": 913114,
      "postDate": "2020-07-03T01:58:46.333Z",
      "rawMarkdown": "",
      "votes": -1,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 913114,
      "author_name": "",
      "author_url": "",
      "post_date": "2020-07-03T01:58:46.333000",
      "content": "",
      "votes": -1,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "912437": "This is my first Kaggle competition and my first one using computer vision (I only worked on tabular data previously) and I'm wondering whether it's possible to get a better score than 0.92 using only a given fraction of the images ? \n\nI'm asking this because the data is very large and it's infeasible to train CNN like EffNet B(0-7) and others using all images and Kaggle runtime. I can't use my computer since I don't have a powerful GPU just 4 GB of RAM and I5 for CPU and my internet connection is very slow (2MB/s), so the only way for me to learn and participate in such competitions is using Kaggle Runtime or Colab.  ",
    "913114": ""
  }
}