{
  "id": 223438,
  "title": "How to augment noise in images?",
  "url": "/competitions/bms-molecular-translation/discussion/223438",
  "author_name": "",
  "post_date": "2021-03-03T20:18:17.152863900Z",
  "votes": 7,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi fellow Kagglers,<br>\nAfter going through <a href=\"https://www.kaggle.com/ihelon/molecular-translation-exploratory-data-analysis/notebook\" target=\"_blank\">this</a> amazing EDA notebook by <a href=\"https://www.kaggle.com/ihelon\" target=\"_blank\">@ihelon</a> , it is quite apparent that there is some salt-pepper or random noise along with the molecular structure diagram in most of the images.<br>\nAnd as per the organizers the noise might vary in the test set also. So, I was thinking adding augmented noise to the train-set as well should be a good idea.</p>\n<p>But I do not know how to add noise to images during augmentation step. Any link or resources will be quite helpful.</p>\n<p>I am very new to CV so it might sound as a very naïve question, but I really need to know this.</p>\n<p>Thanks in advance 😊</p>",
  "messages": [
    {
      "id": "1225685",
      "postDate": "03/03/2021 20:18:17",
      "content": "<p>Hi fellow Kagglers,<br>\nAfter going through <a href=\"https://www.kaggle.com/ihelon/molecular-translation-exploratory-data-analysis/notebook\" target=\"_blank\">this</a> amazing EDA notebook by <a href=\"https://www.kaggle.com/ihelon\" target=\"_blank\">@ihelon</a> , it is quite apparent that there is some salt-pepper or random noise along with the molecular structure diagram in most of the images.<br>\nAnd as per the organizers the noise might vary in the test set also. So, I was thinking adding augmented noise to the train-set as well should be a good idea.</p>\n<p>But I do not know how to add noise to images during augmentation step. Any link or resources will be quite helpful.</p>\n<p>I am very new to CV so it might sound as a very naïve question, but I really need to know this.</p>\n<p>Thanks in advance 😊</p>",
      "rawMarkdown": "Hi fellow Kagglers,\nAfter going through [this](https://www.kaggle.com/ihelon/molecular-translation-exploratory-data-analysis/notebook) amazing EDA notebook by @ihelon , it is quite apparent that there is some salt-pepper or random noise along with the molecular structure diagram in most of the images.\nAnd as per the organizers the noise might vary in the test set also. So, I was thinking adding augmented noise to the train-set as well should be a good idea.\n\nBut I do not know how to add noise to images during augmentation step. Any link or resources will be quite helpful.\n\nI am very new to CV so it might sound as a very naïve question, but I really need to know this.\n\nThanks in advance 😊",
      "votes": null
    },
    {
      "id": "1225688",
      "postDate": "03/03/2021 20:26:37",
      "content": "<p><a href=\"https://www.kaggle.com/manabendrarout\" target=\"_blank\">@manabendrarout</a> Hi!<br>\nLook at this library <a href=\"https://albumentations.ai/docs/getting_started/image_augmentation/\" target=\"_blank\">albumentations.ai</a>. I think this library is very easy and useful.<br>\nThe example of using these augmentations you can also look at one of my kernels, for example <a href=\"https://www.kaggle.com/ihelon/lego-minifigures-tensorflow-tutorial\" target=\"_blank\">LEGO Minifigures - TensorFlow Tutorial</a></p>",
      "rawMarkdown": "manabendrarout Hi!\nLook at this library [albumentations.ai](https://albumentations.ai/docs/getting_started/image_augmentation/). I think this library is very easy and useful.\nThe example of using these augmentations you can also look at one of my kernels, for example [LEGO Minifigures - TensorFlow Tutorial](https://www.kaggle.com/ihelon/lego-minifigures-tensorflow-tutorial)",
      "votes": null
    },
    {
      "id": "1225737",
      "postDate": "03/03/2021 21:41:48",
      "content": "<p>I didn't know that albumentations can be used so smoothly with TF 2.x, thanks for share your work.</p>",
      "rawMarkdown": "I didn't know that albumentations can be used so smoothly with TF 2.x, thanks for share your work.",
      "votes": null
    },
    {
      "id": "1225972",
      "postDate": "03/04/2021 05:36:29",
      "content": "<p>Thanks <a href=\"https://www.kaggle.com/ihelon\" target=\"_blank\">@ihelon</a> <br>\nI wasn't aware albumentations had a noise feature as well… Thanks a lot for sharing 😃</p>",
      "rawMarkdown": "Thanks @ihelon \nI wasn't aware albumentations had a noise feature as well... Thanks a lot for sharing 😃",
      "votes": null
    },
    {
      "id": "1233310",
      "postDate": "03/10/2021 09:47:02",
      "content": "<p>For people coming here to see any specific implementations, you can follow the links below for Albumentation's official documentation on each:-</p>\n<ol>\n<li><a href=\"https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.GaussNoise\" target=\"_blank\">Gaaussian Noise</a></li>\n<li><a href=\"https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.ISONoise\" target=\"_blank\">Iso Noise</a></li>\n<li><a href=\"https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.MultiplicativeNoise\" target=\"_blank\">Multiplicative Noise</a></li>\n</ol>",
      "rawMarkdown": "For people coming here to see any specific implementations, you can follow the links below for Albumentation's official documentation on each:-\n1. [Gaaussian Noise](https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.GaussNoise)\n2. [Iso Noise](https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.ISONoise)\n3. [Multiplicative Noise](https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.MultiplicativeNoise)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 1225688,
      "author_name": "ihelon",
      "author_url": "",
      "post_date": "03/03/2021 20:26:37",
      "content": "<p><a href=\"https://www.kaggle.com/manabendrarout\" target=\"_blank\">@manabendrarout</a> Hi!<br>\nLook at this library <a href=\"https://albumentations.ai/docs/getting_started/image_augmentation/\" target=\"_blank\">albumentations.ai</a>. I think this library is very easy and useful.<br>\nThe example of using these augmentations you can also look at one of my kernels, for example <a href=\"https://www.kaggle.com/ihelon/lego-minifigures-tensorflow-tutorial\" target=\"_blank\">LEGO Minifigures - TensorFlow Tutorial</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 1225737,
          "author_name": "hiramcho",
          "author_url": "",
          "post_date": "03/03/2021 21:41:48",
          "content": "<p>I didn't know that albumentations can be used so smoothly with TF 2.x, thanks for share your work.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 1225972,
          "author_name": "manabendrarout",
          "author_url": "",
          "post_date": "03/04/2021 05:36:29",
          "content": "<p>Thanks <a href=\"https://www.kaggle.com/ihelon\" target=\"_blank\">@ihelon</a> <br>\nI wasn't aware albumentations had a noise feature as well… Thanks a lot for sharing 😃</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 1233310,
      "author_name": "manabendrarout",
      "author_url": "",
      "post_date": "03/10/2021 09:47:02",
      "content": "<p>For people coming here to see any specific implementations, you can follow the links below for Albumentation's official documentation on each:-</p>\n<ol>\n<li><a href=\"https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.GaussNoise\" target=\"_blank\">Gaaussian Noise</a></li>\n<li><a href=\"https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.ISONoise\" target=\"_blank\">Iso Noise</a></li>\n<li><a href=\"https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.MultiplicativeNoise\" target=\"_blank\">Multiplicative Noise</a></li>\n</ol>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "1225685": "Hi fellow Kagglers,\nAfter going through [this](https://www.kaggle.com/ihelon/molecular-translation-exploratory-data-analysis/notebook) amazing EDA notebook by @ihelon , it is quite apparent that there is some salt-pepper or random noise along with the molecular structure diagram in most of the images.\nAnd as per the organizers the noise might vary in the test set also. So, I was thinking adding augmented noise to the train-set as well should be a good idea.\n\nBut I do not know how to add noise to images during augmentation step. Any link or resources will be quite helpful.\n\nI am very new to CV so it might sound as a very naïve question, but I really need to know this.\n\nThanks in advance 😊",
    "1225688": "manabendrarout Hi!\nLook at this library [albumentations.ai](https://albumentations.ai/docs/getting_started/image_augmentation/). I think this library is very easy and useful.\nThe example of using these augmentations you can also look at one of my kernels, for example [LEGO Minifigures - TensorFlow Tutorial](https://www.kaggle.com/ihelon/lego-minifigures-tensorflow-tutorial)",
    "1225737": "I didn't know that albumentations can be used so smoothly with TF 2.x, thanks for share your work.",
    "1225972": "Thanks @ihelon \nI wasn't aware albumentations had a noise feature as well... Thanks a lot for sharing 😃",
    "1233310": "For people coming here to see any specific implementations, you can follow the links below for Albumentation's official documentation on each:-\n1. [Gaaussian Noise](https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.GaussNoise)\n2. [Iso Noise](https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.ISONoise)\n3. [Multiplicative Noise](https://albumentations.ai/docs/api_reference/augmentations/transforms/#albumentations.augmentations.transforms.MultiplicativeNoise)"
  },
  "source": "meta"
}