{
  "id": 549310,
  "title": "stuck at 0.9187",
  "url": "/competitions/cidaut-ai-fake-scene-classification-2024/discussion/549310",
  "author_name": "",
  "post_date": "2024-12-01T17:55:40.854110100Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I am stuck at 0.9187, I split dataset into 80/20 split, and I am pretty sure I exactly minimize the model even on the test set since I last got 0.00009659818 test loss (BCE) and 100% test accuracy. I trained multiple models and my models always generate exactly the same prediction on the validation set which has a score of 0.9187.</p>\n<p>EDIT: I used a 25% train / 75% test split and still got 100% test accuracy. I have confirmed that my splits are correct and manually reviewed my model predictions.  At 10% train / 90% test I get 97% accuracy.</p>",
  "messages": [
    {
      "id": "3060422",
      "postDate": "12/01/2024 17:55:40",
      "content": "<p>I am stuck at 0.9187, I split dataset into 80/20 split, and I am pretty sure I exactly minimize the model even on the test set since I last got 0.00009659818 test loss (BCE) and 100% test accuracy. I trained multiple models and my models always generate exactly the same prediction on the validation set which has a score of 0.9187.</p>\n<p>EDIT: I used a 25% train / 75% test split and still got 100% test accuracy. I have confirmed that my splits are correct and manually reviewed my model predictions.  At 10% train / 90% test I get 97% accuracy.</p>",
      "rawMarkdown": "I am stuck at 0.9187, I split dataset into 80/20 split, and I am pretty sure I exactly minimize the model even on the test set since I last got 0.00009659818 test loss (BCE) and 100% test accuracy. I trained multiple models and my models always generate exactly the same prediction on the validation set which has a score of 0.9187.\n\nEDIT: I used a 25% train / 75% test split and still got 100% test accuracy. I have confirmed that my splits are correct and manually reviewed my model predictions.  At 10% train / 90% test I get 97% accuracy.",
      "votes": null
    },
    {
      "id": "3060721",
      "postDate": "12/02/2024 03:41:43",
      "content": "<p>hi me too </p>",
      "rawMarkdown": "hi me too",
      "votes": null
    },
    {
      "id": "3065316",
      "postDate": "12/06/2024 17:06:42",
      "content": "<p>try applying filter to the image i used log and hsv filter which helps me to breakthrough but am all so stuck in 0.9325 </p>",
      "rawMarkdown": "try applying filter to the image i used log and hsv filter which helps me to breakthrough but am all so stuck in 0.9325",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 3060721,
      "author_name": "hwkims",
      "author_url": "",
      "post_date": "12/02/2024 03:41:43",
      "content": "<p>hi me too </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 3065316,
      "author_name": "",
      "author_url": "",
      "post_date": "12/06/2024 17:06:42",
      "content": "<p>try applying filter to the image i used log and hsv filter which helps me to breakthrough but am all so stuck in 0.9325 </p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "3060422": "I am stuck at 0.9187, I split dataset into 80/20 split, and I am pretty sure I exactly minimize the model even on the test set since I last got 0.00009659818 test loss (BCE) and 100% test accuracy. I trained multiple models and my models always generate exactly the same prediction on the validation set which has a score of 0.9187.\n\nEDIT: I used a 25% train / 75% test split and still got 100% test accuracy. I have confirmed that my splits are correct and manually reviewed my model predictions.  At 10% train / 90% test I get 97% accuracy.",
    "3060721": "hi me too",
    "3065316": "try applying filter to the image i used log and hsv filter which helps me to breakthrough but am all so stuck in 0.9325"
  },
  "source": "meta"
}