![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F3962172%2F97498b842ea3768069dac87740b348af%2F__results___4_0.png?generation=1590739844400629&alt=media\")
PANDA: train 224x224 cropped & non distorted
\n\n\n\nThe image processing consists of two steps:\n1. Cropping white background\n2. Resizing to square with respect to original proportions
\n\n
And as a result the actual biopsy area is centered without distortion in images.
\n\nI've decided to play with Resnet50, so I will start with its default input shape (224, 224, 3).
", "messages": [ { "id": 866244, "postDate": "2020-05-29T08:18:46.673Z", "content": "PANDA: train 224x224 cropped & non distorted
\n\n\n\nThe image processing consists of two steps:\n1. Cropping white background\n2. Resizing to square with respect to original proportions
\n\n
And as a result the actual biopsy area is centered without distortion in images.
\n\nI've decided to play with Resnet50, so I will start with its default input shape (224, 224, 3).
", "rawMarkdown": "[PANDA: train 224x224 cropped & non distorted](https://www.kaggle.com/raibektussupbekov/panda-train-224x224-cropped-non-distorted)\n\n[The notebook can be utilized to produce train images with another shape on base of another TIFF level.](https://www.kaggle.com/raibektussupbekov/panda-crop-resize-to-square-save-train-data)\n\nThe image processing consists of two steps:\n1. Cropping white background\n2. Resizing to square with respect to original proportions\n\n\n\nAnd as a result the actual biopsy area is centered without distortion in images.\n\nI've decided to play with Resnet50, so I will start with its default input shape (224, 224, 3).", "votes": 3 }, { "id": 866248, "postDate": "2020-05-29T08:20:04.383Z", "content": "Thanks for this new dataset
", "rawMarkdown": "Thanks for this new dataset", "replies": [ { "id": 866250, "postDate": "2020-05-29T08:21:16.713Z", "content": "You're welcome😊
", "rawMarkdown": "You're welcome😊 ", "votes": 1 }, { "id": 866254, "postDate": "2020-05-29T08:23:14.937Z", "content": "224 px images doesn't work so well for the classification task, so instead it may have been better to create a dataset with 512 px images
", "rawMarkdown": "224 px images doesn't work so well for the classification task, so instead it may have been better to create a dataset with 512 px images", "votes": 1 }, { "id": 866260, "postDate": "2020-05-29T08:27:07.417Z", "content": "You can easily do it using my notebook https://www.kaggle.com/raibektussupbekov/panda-crop-resize-to-square-save-train-data😉 \nJust set IMG_SIZE = 512 and , if you want to process higher dimension images, set TIFF_LEVEL = 0.
Thanks for this new dataset
", "votes": 0, "replies": [ { "id": 866250, "author_name": "Raibek", "author_url": "", "post_date": "2020-05-29T08:21:16.713000", "content": "You're welcome😊
", "votes": 1, "replies": [] }, { "id": 866254, "author_name": "Kurian Benoy", "author_url": "", "post_date": "2020-05-29T08:23:14.937000", "content": "224 px images doesn't work so well for the classification task, so instead it may have been better to create a dataset with 512 px images
", "votes": 1, "replies": [] }, { "id": 866260, "author_name": "Raibek", "author_url": "", "post_date": "2020-05-29T08:27:07.417000", "content": "You can easily do it using my notebook https://www.kaggle.com/raibektussupbekov/panda-crop-resize-to-square-save-train-data😉 \nJust set IMG_SIZE = 512 and , if you want to process higher dimension images, set TIFF_LEVEL = 0.