{
  "id": 30818,
  "title": "Missing Labels",
  "url": "/competitions/noaa-fisheries-steller-sea-lion-population-count/discussion/30818",
  "author_name": "",
  "post_date": "2017-03-29T06:12:11.042463200Z",
  "votes": 8,
  "comment_count": 13,
  "views": 0,
  "content": "<p>At first glance I noticed that for many images it was difficult to reconcile the numbers in the CSV with the numbers of dots in the images. Some of the images have missing dots, e.g. image 235 has no red labels. Then there is image 66 which is interesting because it has dotted sea-lions despite none reported in the CSV.</p>",
  "messages": [
    {
      "id": "171236",
      "postDate": "03/29/2017 06:12:11",
      "content": "<p>At first glance I noticed that for many images it was difficult to reconcile the numbers in the CSV with the numbers of dots in the images. Some of the images have missing dots, e.g. image 235 has no red labels. Then there is image 66 which is interesting because it has dotted sea-lions despite none reported in the CSV.</p>",
      "rawMarkdown": "At first glance I noticed that for many images it was difficult to reconcile the numbers in the CSV with the numbers of dots in the images. Some of the images have missing dots, e.g. image 235 has no red labels. Then there is image 66 which is interesting because it has dotted sea-lions despite none reported in the CSV.",
      "votes": null
    },
    {
      "id": "171269",
      "postDate": "03/29/2017 08:48:16",
      "content": "<p>I really like this competition. I hope it is not going to be plagued by poor data. </p>\n\n<p>I have already seen a few concerns, and considering how big the data set is (I'm still downloading), a lot more scrutiny coming</p>",
      "rawMarkdown": "I really like this competition. I hope it is not going to be plagued by poor data. \n\nI have already seen a few concerns, and considering how big the data set is (I'm still downloading), a lot more scrutiny coming",
      "votes": null
    },
    {
      "id": "171321",
      "postDate": "03/29/2017 13:25:34",
      "content": "<p>Just looking through the TrainSmall subset, I can find a bunch of discrepancies. '3' : Train and TrainDotted are totally different; '7' images are rotated 180 degrees; '9' images are way out of alignment. And I can't reconcile the number of dots in '2' (58, counted by hand and code) and the number of sea lions reported in train.csv (60). That's 3 or 4 errors out of 11 test images. Not good odds. </p>",
      "rawMarkdown": "Just looking through the TrainSmall subset, I can find a bunch of discrepancies. '3' : Train and TrainDotted are totally different; '7' images are rotated 180 degrees; '9' images are way out of alignment. And I can't reconcile the number of dots in '2' (58, counted by hand and code) and the number of sea lions reported in train.csv (60). That's 3 or 4 errors out of 11 test images. Not good odds.",
      "votes": null
    },
    {
      "id": "171378",
      "postDate": "03/29/2017 17:33:35",
      "content": "<p>Image 21 (Train vs TrainDotted)</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/171378/6215/21.jpg\" alt=\"enter image description here\" title=\"\"></p>",
      "rawMarkdown": "Image 21 (Train vs TrainDotted)\n\n![enter image description here][1]\n\n\n  [1]: https://kaggle2.blob.core.windows.net/forum-message-attachments/171378/6215/21.jpg",
      "votes": null
    },
    {
      "id": "171431",
      "postDate": "03/29/2017 23:43:08",
      "content": "<p>I also noticed that some dots areoffset from the animals. There are no dot on the animal but there is a dot a little ways away. Also noticed a few missing dots. </p>",
      "rawMarkdown": "I also noticed that some dots areoffset from the animals. There are no dot on the animal but there is a dot a little ways away. Also noticed a few missing dots.",
      "votes": null
    },
    {
      "id": "171435",
      "postDate": "03/30/2017 00:01:04",
      "content": "<p>Thanks for pointing out these discrepancies. We're investigating and will clarify asap.</p>",
      "rawMarkdown": "Thanks for pointing out these discrepancies. We're investigating and will clarify asap.",
      "votes": null
    },
    {
      "id": "171446",
      "postDate": "03/30/2017 01:13:14",
      "content": "<p>A serious alignment problem exits in the TrainDotted and Train dataset in TrainSmall.  Could you please leave us a message when this problem is solved, so that we can re-download the dataset?</p>",
      "rawMarkdown": "A serious alignment problem exits in the TrainDotted and Train dataset in TrainSmall.  Could you please leave us a message when this problem is solved, so that we can re-download the dataset?",
      "votes": null
    },
    {
      "id": "171447",
      "postDate": "03/30/2017 01:22:08",
      "content": "<p>Scanned the whole training dataset and there are obvious discrepancies between the 'Train' and 'TrainDotted' versions of the following images (I used a very crude measure of image similarity, so this list might not be exhaustive):</p>\n\n<p>3, 7,  21, 71, 81, 89, 97, 151, 215, 234, 242, 268, 290, 311, 331, 344, 384, 421, 490, 499, 507, 530, 638, 687,  721, 767, 779, 800, 811, 839, 869, 903, 905, 909, 927</p>\n\n<p>Some of the images appear to be simply flipped or mirrored, while others are of different (and sometimes partially overlapping) fields of view. </p>",
      "rawMarkdown": "Scanned the whole training dataset and there are obvious discrepancies between the 'Train' and 'TrainDotted' versions of the following images (I used a very crude measure of image similarity, so this list might not be exhaustive):\n\n3, 7,  21, 71, 81, 89, 97, 151, 215, 234, 242, 268, 290, 311, 331, 344, 384, 421, 490, 499, 507, 530, 638, 687,  721, 767, 779, 800, 811, 839, 869, 903, 905, 909, 927\n\nSome of the images appear to be simply flipped or mirrored, while others are of different (and sometimes partially overlapping) fields of view.",
      "votes": null
    },
    {
      "id": "171628",
      "postDate": "03/30/2017 20:59:09",
      "content": "<p>Hi everyone, thanks again for flagging these issues. I've posted an update in <a href=\"https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/30895\">another thread</a>.</p>",
      "rawMarkdown": "Hi everyone, thanks again for flagging these issues. I've posted an update in [another thread](https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/30895).",
      "votes": null
    },
    {
      "id": "172006",
      "postDate": "04/01/2017 12:50:17",
      "content": "<p>The same for 2.jpg\nIt contains only 19 blue and 37 brown </p>",
      "rawMarkdown": "The same for 2.jpg\nIt contains only 19 blue and 37 brown",
      "votes": null
    },
    {
      "id": "172259",
      "postDate": "04/03/2017 00:28:31",
      "content": "<p>Some images have shapes which look like sea lions but are unlabeled. Here is an example from image 220.</p>",
      "rawMarkdown": "Some images have shapes which look like sea lions but are unlabeled. Here is an example from image 220.",
      "votes": null
    },
    {
      "id": "172281",
      "postDate": "04/03/2017 02:10:18",
      "content": "<p>I think those are seals and not sea lions. The pups are dark grey too, but pups usually also have a mother next to them. These dark grey animals have no other sea lion nearby, that's why I think they are seals.</p>",
      "rawMarkdown": "I think those are seals and not sea lions. The pups are dark grey too, but pups usually also have a mother next to them. These dark grey animals have no other sea lion nearby, that's why I think they are seals.",
      "votes": null
    },
    {
      "id": "173894",
      "postDate": "04/09/2017 06:15:09",
      "content": "<p>I am enclosing a small section from Image 47, with boxes around two animals that appear to be mislabeled - there is virtually nothing that distinguishes one from the other, and yet, one is labeled an adult_male, and the other a subadult_male? Please clarify.</p>",
      "rawMarkdown": "I am enclosing a small section from Image 47, with boxes around two animals that appear to be mislabeled - there is virtually nothing that distinguishes one from the other, and yet, one is labeled an adult_male, and the other a subadult_male? Please clarify.",
      "votes": null
    },
    {
      "id": "173896",
      "postDate": "04/09/2017 06:19:18",
      "content": "<p>I am enclosing a side-by-side comparison of Image 47 and its dotted version - there appears to be a big section that is blacked out in the dotted image, and which contains a number of animals in the original image. Are these animals in the lower portion of the image to be ignored? A few animals have been boxed in the left image - are these adult sealions?</p>",
      "rawMarkdown": "I am enclosing a side-by-side comparison of Image 47 and its dotted version - there appears to be a big section that is blacked out in the dotted image, and which contains a number of animals in the original image. Are these animals in the lower portion of the image to be ignored? A few animals have been boxed in the left image - are these adult sealions?",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 171269,
      "author_name": "radustoicescu",
      "author_url": "",
      "post_date": "03/29/2017 08:48:16",
      "content": "<p>I really like this competition. I hope it is not going to be plagued by poor data. </p>\n\n<p>I have already seen a few concerns, and considering how big the data set is (I'm still downloading), a lot more scrutiny coming</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171321,
      "author_name": "threeplusone",
      "author_url": "",
      "post_date": "03/29/2017 13:25:34",
      "content": "<p>Just looking through the TrainSmall subset, I can find a bunch of discrepancies. '3' : Train and TrainDotted are totally different; '7' images are rotated 180 degrees; '9' images are way out of alignment. And I can't reconcile the number of dots in '2' (58, counted by hand and code) and the number of sea lions reported in train.csv (60). That's 3 or 4 errors out of 11 test images. Not good odds. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171378,
      "author_name": "zfturbo",
      "author_url": "",
      "post_date": "03/29/2017 17:33:35",
      "content": "<p>Image 21 (Train vs TrainDotted)</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/171378/6215/21.jpg\" alt=\"enter image description here\" title=\"\"></p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171431,
      "author_name": "devinanzelmo",
      "author_url": "",
      "post_date": "03/29/2017 23:43:08",
      "content": "<p>I also noticed that some dots areoffset from the animals. There are no dot on the animal but there is a dot a little ways away. Also noticed a few missing dots. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171435,
      "author_name": "datacanary",
      "author_url": "",
      "post_date": "03/30/2017 00:01:04",
      "content": "<p>Thanks for pointing out these discrepancies. We're investigating and will clarify asap.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171446,
      "author_name": "yuc258",
      "author_url": "",
      "post_date": "03/30/2017 01:13:14",
      "content": "<p>A serious alignment problem exits in the TrainDotted and Train dataset in TrainSmall.  Could you please leave us a message when this problem is solved, so that we can re-download the dataset?</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171447,
      "author_name": "keithchev",
      "author_url": "",
      "post_date": "03/30/2017 01:22:08",
      "content": "<p>Scanned the whole training dataset and there are obvious discrepancies between the 'Train' and 'TrainDotted' versions of the following images (I used a very crude measure of image similarity, so this list might not be exhaustive):</p>\n\n<p>3, 7,  21, 71, 81, 89, 97, 151, 215, 234, 242, 268, 290, 311, 331, 344, 384, 421, 490, 499, 507, 530, 638, 687,  721, 767, 779, 800, 811, 839, 869, 903, 905, 909, 927</p>\n\n<p>Some of the images appear to be simply flipped or mirrored, while others are of different (and sometimes partially overlapping) fields of view. </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 171628,
      "author_name": "datacanary",
      "author_url": "",
      "post_date": "03/30/2017 20:59:09",
      "content": "<p>Hi everyone, thanks again for flagging these issues. I've posted an update in <a href=\"https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/30895\">another thread</a>.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 172006,
      "author_name": "sh1ngg",
      "author_url": "",
      "post_date": "04/01/2017 12:50:17",
      "content": "<p>The same for 2.jpg\nIt contains only 19 blue and 37 brown </p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 172259,
      "author_name": "synapticsugar",
      "author_url": "",
      "post_date": "04/03/2017 00:28:31",
      "content": "<p>Some images have shapes which look like sea lions but are unlabeled. Here is an example from image 220.</p>",
      "votes": null,
      "replies": [
        {
          "id": 172281,
          "author_name": "andraszsom",
          "author_url": "",
          "post_date": "04/03/2017 02:10:18",
          "content": "<p>I think those are seals and not sea lions. The pups are dark grey too, but pups usually also have a mother next to them. These dark grey animals have no other sea lion nearby, that's why I think they are seals.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 173894,
      "author_name": "seswar",
      "author_url": "",
      "post_date": "04/09/2017 06:15:09",
      "content": "<p>I am enclosing a small section from Image 47, with boxes around two animals that appear to be mislabeled - there is virtually nothing that distinguishes one from the other, and yet, one is labeled an adult_male, and the other a subadult_male? Please clarify.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 173896,
      "author_name": "seswar",
      "author_url": "",
      "post_date": "04/09/2017 06:19:18",
      "content": "<p>I am enclosing a side-by-side comparison of Image 47 and its dotted version - there appears to be a big section that is blacked out in the dotted image, and which contains a number of animals in the original image. Are these animals in the lower portion of the image to be ignored? A few animals have been boxed in the left image - are these adult sealions?</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "171236": "At first glance I noticed that for many images it was difficult to reconcile the numbers in the CSV with the numbers of dots in the images. Some of the images have missing dots, e.g. image 235 has no red labels. Then there is image 66 which is interesting because it has dotted sea-lions despite none reported in the CSV.",
    "171269": "I really like this competition. I hope it is not going to be plagued by poor data. \n\nI have already seen a few concerns, and considering how big the data set is (I'm still downloading), a lot more scrutiny coming",
    "171321": "Just looking through the TrainSmall subset, I can find a bunch of discrepancies. '3' : Train and TrainDotted are totally different; '7' images are rotated 180 degrees; '9' images are way out of alignment. And I can't reconcile the number of dots in '2' (58, counted by hand and code) and the number of sea lions reported in train.csv (60). That's 3 or 4 errors out of 11 test images. Not good odds.",
    "171378": "Image 21 (Train vs TrainDotted)\n\n![enter image description here][1]\n\n\n  [1]: https://kaggle2.blob.core.windows.net/forum-message-attachments/171378/6215/21.jpg",
    "171431": "I also noticed that some dots areoffset from the animals. There are no dot on the animal but there is a dot a little ways away. Also noticed a few missing dots.",
    "171435": "Thanks for pointing out these discrepancies. We're investigating and will clarify asap.",
    "171446": "A serious alignment problem exits in the TrainDotted and Train dataset in TrainSmall.  Could you please leave us a message when this problem is solved, so that we can re-download the dataset?",
    "171447": "Scanned the whole training dataset and there are obvious discrepancies between the 'Train' and 'TrainDotted' versions of the following images (I used a very crude measure of image similarity, so this list might not be exhaustive):\n\n3, 7,  21, 71, 81, 89, 97, 151, 215, 234, 242, 268, 290, 311, 331, 344, 384, 421, 490, 499, 507, 530, 638, 687,  721, 767, 779, 800, 811, 839, 869, 903, 905, 909, 927\n\nSome of the images appear to be simply flipped or mirrored, while others are of different (and sometimes partially overlapping) fields of view.",
    "171628": "Hi everyone, thanks again for flagging these issues. I've posted an update in [another thread](https://www.kaggle.com/c/noaa-fisheries-steller-sea-lion-population-count/discussion/30895).",
    "172006": "The same for 2.jpg\nIt contains only 19 blue and 37 brown",
    "172259": "Some images have shapes which look like sea lions but are unlabeled. Here is an example from image 220.",
    "172281": "I think those are seals and not sea lions. The pups are dark grey too, but pups usually also have a mother next to them. These dark grey animals have no other sea lion nearby, that's why I think they are seals.",
    "173894": "I am enclosing a small section from Image 47, with boxes around two animals that appear to be mislabeled - there is virtually nothing that distinguishes one from the other, and yet, one is labeled an adult_male, and the other a subadult_male? Please clarify.",
    "173896": "I am enclosing a side-by-side comparison of Image 47 and its dotted version - there appears to be a big section that is blacked out in the dotted image, and which contains a number of animals in the original image. Are these animals in the lower portion of the image to be ignored? A few animals have been boxed in the left image - are these adult sealions?"
  },
  "source": "meta"
}