![\"\"](\"https://i.ibb.co/DL2m8FJ/robga-examples.png\")
This is a list of badly formed training images. You may wish to remove them, or save good portions of the large images.\n046586a, 1588d4c, 1e40a05, 41f92e5, 449b792, 563fc48, 8bd81ce, b092cc1, c0306e5, c26c635, e04fea3, e5f2f24, eda52f2, fa645da
\n\nThere are other images with problems that in my experience aren't bad enough to remove from the training set.\n- join issues as in d821c94\n- half images as in 24884e7, there are quite a few of these\n- colour abberations as in d744e88\n- scan line striations as in d9989e9, these are commonly seen in bright areas of satellite images
\n\nFor completion, these are bad test images. Not much you can do about them.
\n\n18c7c27, 3cb4ce5, 25c5403, 43d43a8, bf4ddbb, 4375f9e, 638185f, d1378d2
", "messages": [ { "id": "605660", "postDate": "08/22/2019 15:41:10", "content": "This is a list of badly formed training images. You may wish to remove them, or save good portions of the large images.\n046586a, 1588d4c, 1e40a05, 41f92e5, 449b792, 563fc48, 8bd81ce, b092cc1, c0306e5, c26c635, e04fea3, e5f2f24, eda52f2, fa645da
\n\nThere are other images with problems that in my experience aren't bad enough to remove from the training set.\n- join issues as in d821c94\n- half images as in 24884e7, there are quite a few of these\n- colour abberations as in d744e88\n- scan line striations as in d9989e9, these are commonly seen in bright areas of satellite images
\n\nFor completion, these are bad test images. Not much you can do about them.
\n\n18c7c27, 3cb4ce5, 25c5403, 43d43a8, bf4ddbb, 4375f9e, 638185f, d1378d2
", "rawMarkdown": "This is a list of badly formed training images. You may wish to remove them, or save good portions of the large images.\n046586a, 1588d4c, 1e40a05, 41f92e5, 449b792, 563fc48, 8bd81ce, b092cc1, c0306e5, c26c635, e04fea3, e5f2f24, eda52f2, fa645da\n\nThere are other images with problems that in my experience aren't bad enough to remove from the training set.\n- join issues as in d821c94\n- half images as in 24884e7, there are quite a few of these\n- colour abberations as in d744e88\n- scan line striations as in d9989e9, these are commonly seen in bright areas of satellite images\n\nFor completion, these are bad test images. Not much you can do about them. \n\n18c7c27, 3cb4ce5, 25c5403, 43d43a8, bf4ddbb, 4375f9e, 638185f, d1378d2", "votes": null }, { "id": "606087", "postDate": "08/23/2019 07:10:28", "content": "Thanks for the list!\nI don't see how scraping data would help this competition. Do you mean getting more satellite images, manually drawing the bounding rectangles and creating the segmentation with them, then finally using that as extra training data?
", "rawMarkdown": "Thanks for the list!\nI don't see how scraping data would help this competition. Do you mean getting more satellite images, manually drawing the bounding rectangles and creating the segmentation with them, then finally using that as extra training data?", "votes": null }, { "id": "606102", "postDate": "08/23/2019 07:36:07", "content": "", "rawMarkdown": "", "votes": null }, { "id": "606695", "postDate": "08/23/2019 23:21:08", "content": "I didn't know about noisy labels. Thank you for clarifying.
", "rawMarkdown": "I didn't know about noisy labels. Thank you for clarifying.", "votes": null }, { "id": "627864", "postDate": "09/16/2019 13:46:30", "content": "Thank you for posting this list, it has saved me some time and effort!\nI had just found one of these and checked the discussion board to see if anyone else had come across similar issues.
", "rawMarkdown": "Thank you for posting this list, it has saved me some time and effort!\nI had just found one of these and checked the discussion board to see if anyone else had come across similar issues.", "votes": null }, { "id": "629785", "postDate": "09/19/2019 07:04:16", "content": "For anyone who is interested. These are examples on @robga bad list (and not-too-bad ones :)
\n\n
@robga Thank you for sharing. I found '38185f' in the list of bad test images might be typo. I think it should be '638185f'.
", "rawMarkdown": "robga Thank you for sharing. I found '38185f' in the list of bad test images might be typo. I think it should be '638185f'.", "votes": null } ], "comments": [ { "id": 606087, "author_name": "renanod", "author_url": "", "post_date": "08/23/2019 07:10:28", "content": "Thanks for the list!\nI don't see how scraping data would help this competition. Do you mean getting more satellite images, manually drawing the bounding rectangles and creating the segmentation with them, then finally using that as extra training data?
", "votes": null, "replies": [ { "id": 606102, "author_name": "robga", "author_url": "", "post_date": "08/23/2019 07:36:07", "content": "", "votes": null, "replies": [] }, { "id": 606695, "author_name": "renanod", "author_url": "", "post_date": "08/23/2019 23:21:08", "content": "I didn't know about noisy labels. Thank you for clarifying.
", "votes": null, "replies": [] } ] }, { "id": 627864, "author_name": "anthony35813", "author_url": "", "post_date": "09/16/2019 13:46:30", "content": "Thank you for posting this list, it has saved me some time and effort!\nI had just found one of these and checked the discussion board to see if anyone else had come across similar issues.
", "votes": null, "replies": [] }, { "id": 629785, "author_name": "ratthachat", "author_url": "", "post_date": "09/19/2019 07:04:16", "content": "For anyone who is interested. These are examples on @robga bad list (and not-too-bad ones :)
\n\n
@robga Thank you for sharing. I found '38185f' in the list of bad test images might be typo. I think it should be '638185f'.
", "votes": null, "replies": [] } ], "raw_markdown_by_id": { "605660": "This is a list of badly formed training images. You may wish to remove them, or save good portions of the large images.\n046586a, 1588d4c, 1e40a05, 41f92e5, 449b792, 563fc48, 8bd81ce, b092cc1, c0306e5, c26c635, e04fea3, e5f2f24, eda52f2, fa645da\n\nThere are other images with problems that in my experience aren't bad enough to remove from the training set.\n- join issues as in d821c94\n- half images as in 24884e7, there are quite a few of these\n- colour abberations as in d744e88\n- scan line striations as in d9989e9, these are commonly seen in bright areas of satellite images\n\nFor completion, these are bad test images. Not much you can do about them. \n\n18c7c27, 3cb4ce5, 25c5403, 43d43a8, bf4ddbb, 4375f9e, 638185f, d1378d2", "606087": "Thanks for the list!\nI don't see how scraping data would help this competition. Do you mean getting more satellite images, manually drawing the bounding rectangles and creating the segmentation with them, then finally using that as extra training data?", "606102": "", "606695": "I didn't know about noisy labels. Thank you for clarifying.", "627864": "Thank you for posting this list, it has saved me some time and effort!\nI had just found one of these and checked the discussion board to see if anyone else had come across similar issues.", "629785": "For anyone who is interested. These are examples on @robga bad list (and not-too-bad ones :)\n\n", "629844": "robga Thank you for sharing. I found '38185f' in the list of bad test images might be typo. I think it should be '638185f'." }, "source": "meta" }