![\"\"](\"https://apps.lucidcentral.org/ppp/images/entities/cassava_brown_streak_disease_439/thumbs/cbsd_root_rot_sml.jpg\")
Hello everyone , from reading other threads in the discussion it is fairly clear that the data contains a lot of noisy images as well as mislabeled data .
\nAfter analyzing the arguments put forward I do verify that there are mislabeled images and noise present in the data , but there are some images in Cassava Brown Streak Disease Segment like fig. below which might appear weird and seem like noise but are fairly accurate
\n
There are actually two types of Images present in this segment one is of the leaves/plant and other is of the Tubular Roots of the Plant . You see the roots are also the source of food when it comes to cassava and can be fairly indicative of a disease condition and hence the image llike above are fairly valid
More can be read from here
\nI have clustered all those images and also the mislabeled images in my recent notebook here. This notebook also contains more interesting insights and there are more to come
\nHope this helps
\nThanks for Reading
Hello everyone , from reading other threads in the discussion it is fairly clear that the data contains a lot of noisy images as well as mislabeled data .
\nAfter analyzing the arguments put forward I do verify that there are mislabeled images and noise present in the data , but there are some images in Cassava Brown Streak Disease Segment like fig. below which might appear weird and seem like noise but are fairly accurate
\n
There are actually two types of Images present in this segment one is of the leaves/plant and other is of the Tubular Roots of the Plant . You see the roots are also the source of food when it comes to cassava and can be fairly indicative of a disease condition and hence the image llike above are fairly valid
More can be read from here
\nI have clustered all those images and also the mislabeled images in my recent notebook here. This notebook also contains more interesting insights and there are more to come
\nHope this helps
\nThanks for Reading
When I read your notebook, I feel I'm reading a survey article. Great EDA. -)
", "rawMarkdown": "When I read your notebook, I feel I'm reading a survey article. Great EDA. -)", "votes": null }, { "id": "1096898", "postDate": "11/30/2020 21:27:13", "content": "Hello, @tanulsingh077! Clustering is very interesting method to find these outliers. One famous question: did you try to drop these \"potatoes\" from dataset and try to learn without it? Is it improves a total score?
", "rawMarkdown": "Hello, @tanulsingh077! Clustering is very interesting method to find these outliers. One famous question: did you try to drop these \"potatoes\" from dataset and try to learn without it? Is it improves a total score?", "votes": null }, { "id": "1097450", "postDate": "12/01/2020 04:03:02", "content": "Thanks for the kind words @ipythonx
\nNew and more interesting Finds coming today
Hi @khlevnov as mentioned the potatoes are actually Roots of the Cassava Tree and play a major role in detection of disease and hence they are absolutely valid Images . No one should remove them
", "rawMarkdown": "Hi @khlevnov as mentioned the potatoes are actually Roots of the Cassava Tree and play a major role in detection of disease and hence they are absolutely valid Images . No one should remove them", "votes": null } ], "comments": [ { "id": 1096492, "author_name": "ipythonx", "author_url": "", "post_date": "11/30/2020 14:32:57", "content": "When I read your notebook, I feel I'm reading a survey article. Great EDA. -)
", "votes": null, "replies": [ { "id": 1097450, "author_name": "tanulsingh077", "author_url": "", "post_date": "12/01/2020 04:03:02", "content": "Thanks for the kind words @ipythonx
\nNew and more interesting Finds coming today
Hello, @tanulsingh077! Clustering is very interesting method to find these outliers. One famous question: did you try to drop these \"potatoes\" from dataset and try to learn without it? Is it improves a total score?
", "votes": null, "replies": [ { "id": 1097451, "author_name": "tanulsingh077", "author_url": "", "post_date": "12/01/2020 04:04:21", "content": "Hi @khlevnov as mentioned the potatoes are actually Roots of the Cassava Tree and play a major role in detection of disease and hence they are absolutely valid Images . No one should remove them
", "votes": null, "replies": [] } ] } ], "raw_markdown_by_id": { "1096309": "Hello everyone , from reading other threads in the discussion it is fairly clear that the data contains a lot of noisy images as well as mislabeled data . \n\nAfter analyzing the arguments put forward I do verify that there are mislabeled images and noise present in the data , but there are some images in Cassava Brown Streak Disease Segment like fig. below which might appear weird and seem like noise but are fairly accurate \n\n\n\n`There are actually two types of Images present in this segment one is of the leaves/plant and other is of the Tubular Roots of the Plant . You see the roots are also the source of food when it comes to cassava and can be fairly indicative of a disease condition and hence the image llike above are fairly valid`\n\nMore can be read from [here] (https://www.apsnet.org/edcenter/apsnetfeatures/Pages/cassava.aspx)\n\nI have clustered all those images and also the mislabeled images in my recent notebook [here](https://www.kaggle.com/tanulsingh077/how-to-become-leaf-doctor-with-deep-learning). This notebook also contains more interesting insights and there are more to come \n\nHope this helps \nThanks for Reading", "1096492": "When I read your notebook, I feel I'm reading a survey article. Great EDA. -)", "1096898": "Hello, @tanulsingh077! Clustering is very interesting method to find these outliers. One famous question: did you try to drop these \"potatoes\" from dataset and try to learn without it? Is it improves a total score?", "1097450": "Thanks for the kind words @ipythonx \nNew and more interesting Finds coming today", "1097451": "Hi @khlevnov as mentioned the potatoes are actually Roots of the Cassava Tree and play a major role in detection of disease and hence they are absolutely valid Images . No one should remove them" }, "source": "meta" }