{
  "id": 379413,
  "title": "ImageDataLoaders",
  "url": "/competitions/rsna-breast-cancer-detection/discussion/379413",
  "author_name": "michael scheinfeild",
  "post_date": "2023-01-19T13:47:16.996000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>hi i saw submission with small data set and not clear for me the data loader loads image and resize to <br>\n166<em>166 but the augmentation size 128</em>128 isnt the sizes should be the same  ? , here is the code  :</p>\n<pre><code>dls = ImageDataLoaders.from_path_func(\n                                small_image_path,\n                                fnames=fn,\n                                label_func=get_y_label,\n                                valid_pct=,\n                                seed=,\n                                item_tfms=Resize(****, method=),\n                                batch_tfms=aug_transforms(**size=**, min_scale=) \n                                )\n</code></pre>",
  "messages": [
    {
      "id": 2106997,
      "postDate": "2023-01-19T13:47:16.997Z",
      "content": "<p>hi i saw submission with small data set and not clear for me the data loader loads image and resize to <br>\n166<em>166 but the augmentation size 128</em>128 isnt the sizes should be the same  ? , here is the code  :</p>\n<pre><code>dls = ImageDataLoaders.from_path_func(\n                                small_image_path,\n                                fnames=fn,\n                                label_func=get_y_label,\n                                valid_pct=,\n                                seed=,\n                                item_tfms=Resize(****, method=),\n                                batch_tfms=aug_transforms(**size=**, min_scale=) \n                                )\n</code></pre>",
      "rawMarkdown": "hi i saw submission with small data set and not clear for me the data loader loads image and resize to \n166*166 but the augmentation size 128*128 isnt the sizes should be the same  ? , here is the code  :\n```python\ndls = ImageDataLoaders.from_path_func(\n                                small_image_path,# path of the image files\n                                fnames=fn,# file names\n                                label_func=get_y_label,# function to get the label\n                                valid_pct=0.2,# percentage of validation data\n                                seed=42,# now seed to get repetitive result\n                                item_tfms=Resize(**166**, method='squish'),# resize image using squash method\n                                batch_tfms=aug_transforms(**size=128**, min_scale=0.75) # in Gpu batch augmenation\n                                )\n```"
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "2106997": "hi i saw submission with small data set and not clear for me the data loader loads image and resize to \n166*166 but the augmentation size 128*128 isnt the sizes should be the same  ? , here is the code  :\n```python\ndls = ImageDataLoaders.from_path_func(\n                                small_image_path,# path of the image files\n                                fnames=fn,# file names\n                                label_func=get_y_label,# function to get the label\n                                valid_pct=0.2,# percentage of validation data\n                                seed=42,# now seed to get repetitive result\n                                item_tfms=Resize(**166**, method='squish'),# resize image using squash method\n                                batch_tfms=aug_transforms(**size=128**, min_scale=0.75) # in Gpu batch augmenation\n                                )\n```"
  }
}