{
  "id": 202580,
  "title": "Augmentation on Validation Images",
  "url": "/competitions/cassava-leaf-disease-classification/discussion/202580",
  "author_name": "",
  "post_date": "2020-12-10T21:22:50.386615300Z",
  "votes": 1,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Hi,</p>\n<p>Please don't mind me asking, I am just starting out and have some doubts. I was following this <a href=\"https://www.kaggle.com/harveenchadha/efficientnetb3-keras-tf2-baseline-training\" target=\"_blank\">kernel</a> to start building my model journey.</p>\n<p>It was kind of quite clear on the steps the author followed, just now I read another <a href=\"https://www.kaggle.com/lavanyask/cassava-leaf-disease-starter-notebook\" target=\"_blank\">kernel</a> , it is kind of similar but there is one difference. This kernel uses Augmentation on validation set as well. </p>\n<p>I am a bit confused, since the latter has more upvotes so does that means that we must do augmentation on the validation set as well?</p>",
  "messages": [
    {
      "id": "1108647",
      "postDate": "12/10/2020 21:22:50",
      "content": "<p>Hi,</p>\n<p>Please don't mind me asking, I am just starting out and have some doubts. I was following this <a href=\"https://www.kaggle.com/harveenchadha/efficientnetb3-keras-tf2-baseline-training\" target=\"_blank\">kernel</a> to start building my model journey.</p>\n<p>It was kind of quite clear on the steps the author followed, just now I read another <a href=\"https://www.kaggle.com/lavanyask/cassava-leaf-disease-starter-notebook\" target=\"_blank\">kernel</a> , it is kind of similar but there is one difference. This kernel uses Augmentation on validation set as well. </p>\n<p>I am a bit confused, since the latter has more upvotes so does that means that we must do augmentation on the validation set as well?</p>",
      "rawMarkdown": "Hi,\n\nPlease don't mind me asking, I am just starting out and have some doubts. I was following this [kernel](https://www.kaggle.com/harveenchadha/efficientnetb3-keras-tf2-baseline-training) to start building my model journey.\n\n It was kind of quite clear on the steps the author followed, just now I read another [kernel](https://www.kaggle.com/lavanyask/cassava-leaf-disease-starter-notebook) , it is kind of similar but there is one difference. This kernel uses Augmentation on validation set as well. \n\nI am a bit confused, since the latter has more upvotes so does that means that we must do augmentation on the validation set as well?",
      "votes": null
    },
    {
      "id": "1109551",
      "postDate": "12/11/2020 20:08:31",
      "content": "<p>No, you dont have to use augmentation on the validset, normally its not done like that.</p>\n<p>There is a technique called TTA (test time augmentation) while interference you are using augmentations on the testset and average all results per image together, this is done because you are also training the model with augmentations.</p>\n<p>Here an explanation: <a href=\"https://www.kaggle.com/andrewkh/test-time-augmentation-tta-worth-it\" target=\"_blank\">https://www.kaggle.com/andrewkh/test-time-augmentation-tta-worth-it</a></p>",
      "rawMarkdown": "No, you dont have to use augmentation on the validset, normally its not done like that.\n\nThere is a technique called TTA (test time augmentation) while interference you are using augmentations on the testset and average all results per image together, this is done because you are also training the model with augmentations.\n\nHere an explanation: https://www.kaggle.com/andrewkh/test-time-augmentation-tta-worth-it",
      "votes": null
    },
    {
      "id": "1113388",
      "postDate": "12/15/2020 12:10:13",
      "content": "<p>I don't think augmentation on validation set should be done unless you have the idea of test time augmentation. All the best!</p>",
      "rawMarkdown": "I don't think augmentation on validation set should be done unless you have the idea of test time augmentation. All the best!",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1109551,
      "author_name": "aliabdin1",
      "author_url": "",
      "post_date": "12/11/2020 20:08:31",
      "content": "<p>No, you dont have to use augmentation on the validset, normally its not done like that.</p>\n<p>There is a technique called TTA (test time augmentation) while interference you are using augmentations on the testset and average all results per image together, this is done because you are also training the model with augmentations.</p>\n<p>Here an explanation: <a href=\"https://www.kaggle.com/andrewkh/test-time-augmentation-tta-worth-it\" target=\"_blank\">https://www.kaggle.com/andrewkh/test-time-augmentation-tta-worth-it</a></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 1113388,
      "author_name": "harveenchadha",
      "author_url": "",
      "post_date": "12/15/2020 12:10:13",
      "content": "<p>I don't think augmentation on validation set should be done unless you have the idea of test time augmentation. All the best!</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1108647": "Hi,\n\nPlease don't mind me asking, I am just starting out and have some doubts. I was following this [kernel](https://www.kaggle.com/harveenchadha/efficientnetb3-keras-tf2-baseline-training) to start building my model journey.\n\n It was kind of quite clear on the steps the author followed, just now I read another [kernel](https://www.kaggle.com/lavanyask/cassava-leaf-disease-starter-notebook) , it is kind of similar but there is one difference. This kernel uses Augmentation on validation set as well. \n\nI am a bit confused, since the latter has more upvotes so does that means that we must do augmentation on the validation set as well?",
    "1109551": "No, you dont have to use augmentation on the validset, normally its not done like that.\n\nThere is a technique called TTA (test time augmentation) while interference you are using augmentations on the testset and average all results per image together, this is done because you are also training the model with augmentations.\n\nHere an explanation: https://www.kaggle.com/andrewkh/test-time-augmentation-tta-worth-it",
    "1113388": "I don't think augmentation on validation set should be done unless you have the idea of test time augmentation. All the best!"
  },
  "source": "meta"
}