![\"\"](\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1684129%2F514cd1f670b1d871769cd2bf4a3b4cc4%2Fmelanoma.PNG?generation=1596391292720314&alt=media\")
\nPlease have a look into the Description of the competition.. It will clear your doubts.Hi everyone. I am just starting up with this competition. Before diving into the modelling bit, I was just doing some data analysis and I found that for a lot of patients in the training dataset, the labels are not consistent for all images of that patient. Is that supposed to be that way or am I missing something here?
\n\nThanks!
", "messages": [ { "id": "955179", "postDate": "08/02/2020 12:21:21", "content": "Hi everyone. I am just starting up with this competition. Before diving into the modelling bit, I was just doing some data analysis and I found that for a lot of patients in the training dataset, the labels are not consistent for all images of that patient. Is that supposed to be that way or am I missing something here?
\n\nThanks!
", "rawMarkdown": "Hi everyone. I am just starting up with this competition. Before diving into the modelling bit, I was just doing some data analysis and I found that for a lot of patients in the training dataset, the labels are not consistent for all images of that patient. Is that supposed to be that way or am I missing something here? \n\nThanks!", "votes": null }, { "id": "955182", "postDate": "08/02/2020 12:23:47", "content": "how are sure that the labels are inconsistent @themlenthusiast ?
", "rawMarkdown": "how are sure that the labels are inconsistent @themlenthusiast ?", "votes": null }, { "id": "955580", "postDate": "08/02/2020 18:02:40", "content": "\nPlease have a look into the Description of the competition.. It will clear your doubts.
We are predicting on an image level not a patient level.
\n\nOne patient can have multiple images and some images can be malignant of benignant depending of the mole. Perhaps thats why you are seeing this \"inconsistencies\" but they are not truly inconsistencies.
", "rawMarkdown": "We are predicting on an image level not a patient level. \n\nOne patient can have multiple images and some images can be malignant of benignant depending of the mole. Perhaps thats why you are seeing this \"inconsistencies\" but they are not truly inconsistencies.", "votes": null }, { "id": "957295", "postDate": "08/04/2020 07:50:24", "content": "Some patients with multiple images may have malignant and non-malignant lesions. That is correct.
", "rawMarkdown": "Some patients with multiple images may have malignant and non-malignant lesions. That is correct.", "votes": null }, { "id": "964485", "postDate": "08/09/2020 23:53:51", "content": "I just looked at the labels corresponding to all the images belonging to one patient. I was expecting that all of them should be the same (and this consistency, if it had been, could have been incorporated into the modelling). The description did not make it clear enough to me but now it does.
", "rawMarkdown": "I just looked at the labels corresponding to all the images belonging to one patient. I was expecting that all of them should be the same (and this consistency, if it had been, could have been incorporated into the modelling). The description did not make it clear enough to me but now it does.", "votes": null }, { "id": "964486", "postDate": "08/09/2020 23:54:59", "content": "Thanks, Shivam. When I re-read it initially, it didn't make this clear for me. But after having spent some time understanding the terminology and the data, this makes it clear now. Thanks!
", "rawMarkdown": "Thanks, Shivam. When I re-read it initially, it didn't make this clear for me. But after having spent some time understanding the terminology and the data, this makes it clear now. Thanks!", "votes": null }, { "id": "964488", "postDate": "08/09/2020 23:55:39", "content": "Makes sense. I didn't understand the concept of moles first. But now, it's clear. Thanks!
", "rawMarkdown": "Makes sense. I didn't understand the concept of moles first. But now, it's clear. Thanks!", "votes": null } ], "comments": [ { "id": 955182, "author_name": "msharuk589", "author_url": "", "post_date": "08/02/2020 12:23:47", "content": "how are sure that the labels are inconsistent @themlenthusiast ?
", "votes": null, "replies": [ { "id": 964485, "author_name": "themlenthusiast", "author_url": "", "post_date": "08/09/2020 23:53:51", "content": "I just looked at the labels corresponding to all the images belonging to one patient. I was expecting that all of them should be the same (and this consistency, if it had been, could have been incorporated into the modelling). The description did not make it clear enough to me but now it does.
", "votes": null, "replies": [] } ] }, { "id": 955580, "author_name": "shivamcyborg", "author_url": "", "post_date": "08/02/2020 18:02:40", "content": "\nPlease have a look into the Description of the competition.. It will clear your doubts.
Thanks, Shivam. When I re-read it initially, it didn't make this clear for me. But after having spent some time understanding the terminology and the data, this makes it clear now. Thanks!
", "votes": null, "replies": [] } ] }, { "id": 956619, "author_name": "santiviquez", "author_url": "", "post_date": "08/03/2020 16:20:57", "content": "We are predicting on an image level not a patient level.
\n\nOne patient can have multiple images and some images can be malignant of benignant depending of the mole. Perhaps thats why you are seeing this \"inconsistencies\" but they are not truly inconsistencies.
", "votes": null, "replies": [ { "id": 964488, "author_name": "themlenthusiast", "author_url": "", "post_date": "08/09/2020 23:55:39", "content": "Makes sense. I didn't understand the concept of moles first. But now, it's clear. Thanks!
", "votes": null, "replies": [] } ] }, { "id": 957295, "author_name": "jsyphil", "author_url": "", "post_date": "08/04/2020 07:50:24", "content": "Some patients with multiple images may have malignant and non-malignant lesions. That is correct.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "955179": "Hi everyone. I am just starting up with this competition. Before diving into the modelling bit, I was just doing some data analysis and I found that for a lot of patients in the training dataset, the labels are not consistent for all images of that patient. Is that supposed to be that way or am I missing something here? \n\nThanks!", "955182": "how are sure that the labels are inconsistent @themlenthusiast ?", "955580": "\nPlease have a look into the Description of the competition.. It will clear your doubts.", "956619": "We are predicting on an image level not a patient level. \n\nOne patient can have multiple images and some images can be malignant of benignant depending of the mole. Perhaps thats why you are seeing this \"inconsistencies\" but they are not truly inconsistencies.", "957295": "Some patients with multiple images may have malignant and non-malignant lesions. That is correct.", "964485": "I just looked at the labels corresponding to all the images belonging to one patient. I was expecting that all of them should be the same (and this consistency, if it had been, could have been incorporated into the modelling). The description did not make it clear enough to me but now it does.", "964486": "Thanks, Shivam. When I re-read it initially, it didn't make this clear for me. But after having spent some time understanding the terminology and the data, this makes it clear now. Thanks!", "964488": "Makes sense. I didn't understand the concept of moles first. But now, it's clear. Thanks!" }, "source": "meta" }