{
  "id": 15192,
  "title": "Minimum Image Size",
  "url": "/competitions/diabetic-retinopathy-detection/discussion/15192",
  "author_name": "clustifier",
  "post_date": "2015-07-12T18:58:21.483000",
  "votes": 1,
  "comment_count": 7,
  "views": 1652,
  "content": "<p>Hi all,</p>\n\n<p>What do you think is the minimum image needed to get meaningful results?\nUntil now I didn't managed to build any reasonable model because of storage limitation on my laptop.</p>\n\n<p>Thanks,\nC</p>",
  "messages": [
    {
      "id": 84355,
      "postDate": "2015-07-14T07:28:29.310Z",
      "content": "<p>I got above 0.70 with just 120x120. So there is still a lot of information in the low level features.</p>",
      "rawMarkdown": "I got above 0.70 with just 120x120. So there is still a lot of information in the low level features.",
      "votes": 5
    },
    {
      "id": 84359,
      "postDate": "2015-07-14T07:42:12.660Z",
      "content": "<p>We got just a tiny fraction above 0.7 with 128x128. 96x96 and 256x256 got us slightly worse results (~ 0.67)</p>",
      "rawMarkdown": "We got just a tiny fraction above 0.7 with 128x128. 96x96 and 256x256 got us slightly worse results (~ 0.67)",
      "votes": 3
    },
    {
      "id": 84325,
      "postDate": "2015-07-14T01:50:19.780Z",
      "content": "<p>I was able to get 0.18 using 48x48, 0.30 using 96x96, with no data augmentation.</p>",
      "rawMarkdown": "I was able to get 0.18 using 48x48, 0.30 using 96x96, with no data augmentation.",
      "votes": 3
    },
    {
      "id": 84231,
      "postDate": "2015-07-12T18:58:21.483Z",
      "content": "<p>Hi all,</p>\n\n<p>What do you think is the minimum image needed to get meaningful results?\nUntil now I didn't managed to build any reasonable model because of storage limitation on my laptop.</p>\n\n<p>Thanks,\nC</p>",
      "rawMarkdown": "Hi all,\r\n\r\nWhat do you think is the minimum image needed to get meaningful results?\r\nUntil now I didn't managed to build any reasonable model because of storage limitation on my laptop.\r\n\r\nThanks,\r\nC\r\n ",
      "votes": 1
    },
    {
      "id": 84263,
      "postDate": "2015-07-13T10:04:50.487Z",
      "content": "<p>@Halla Yang Could you tell me, what size you use?</p>",
      "rawMarkdown": "@Halla Yang Could you tell me, what size you use?\r\n",
      "votes": -2
    },
    {
      "id": 84358,
      "postDate": "2015-07-14T07:38:40.260Z",
      "content": "<p>Hey guys! Could somebody please kindly explain me, why I got down votes to my last question, while more other ones share their thoughts regarding image size? Thank you in advance.</p>",
      "rawMarkdown": "Hey guys! Could somebody please kindly explain me, why I got down votes to my last question, while more other ones share their thoughts regarding image size? Thank you in advance."
    },
    {
      "id": 84253,
      "postDate": "2015-07-13T05:41:16.400Z",
      "content": "<p>thank you!</p>\n\n<p>looks it's too much for my laptop.</p>",
      "rawMarkdown": "thank you!\r\n\r\nlooks it's too much for my laptop.\r\n"
    },
    {
      "id": 84232,
      "postDate": "2015-07-12T19:10:24.447Z",
      "rawMarkdown": "",
      "votes": 3,
      "isDeleted": true
    }
  ],
  "comments": [
    {
      "id": 84355,
      "author_name": "Jeffrey",
      "author_url": "",
      "post_date": "2015-07-14T07:28:29.310000",
      "content": "<p>I got above 0.70 with just 120x120. So there is still a lot of information in the low level features.</p>",
      "votes": 5,
      "replies": []
    },
    {
      "id": 84359,
      "author_name": "sovcharenko",
      "author_url": "",
      "post_date": "2015-07-14T07:42:12.660000",
      "content": "<p>We got just a tiny fraction above 0.7 with 128x128. 96x96 and 256x256 got us slightly worse results (~ 0.67)</p>",
      "votes": 3,
      "replies": []
    },
    {
      "id": 84325,
      "author_name": "James King",
      "author_url": "",
      "post_date": "2015-07-14T01:50:19.780000",
      "content": "<p>I was able to get 0.18 using 48x48, 0.30 using 96x96, with no data augmentation.</p>",
      "votes": 3,
      "replies": []
    },
    {
      "id": 84263,
      "author_name": "",
      "author_url": "",
      "post_date": "2015-07-13T10:04:50.487000",
      "content": "<p>@Halla Yang Could you tell me, what size you use?</p>",
      "votes": -2,
      "replies": []
    },
    {
      "id": 84358,
      "author_name": "",
      "author_url": "",
      "post_date": "2015-07-14T07:38:40.260000",
      "content": "<p>Hey guys! Could somebody please kindly explain me, why I got down votes to my last question, while more other ones share their thoughts regarding image size? Thank you in advance.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 84253,
      "author_name": "clustifier",
      "author_url": "",
      "post_date": "2015-07-13T05:41:16.400000",
      "content": "<p>thank you!</p>\n\n<p>looks it's too much for my laptop.</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 84232,
      "author_name": "",
      "author_url": "",
      "post_date": "2015-07-12T19:10:24.447000",
      "content": "",
      "votes": 3,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "84355": "I got above 0.70 with just 120x120. So there is still a lot of information in the low level features.",
    "84359": "We got just a tiny fraction above 0.7 with 128x128. 96x96 and 256x256 got us slightly worse results (~ 0.67)",
    "84325": "I was able to get 0.18 using 48x48, 0.30 using 96x96, with no data augmentation.",
    "84231": "Hi all,\r\n\r\nWhat do you think is the minimum image needed to get meaningful results?\r\nUntil now I didn't managed to build any reasonable model because of storage limitation on my laptop.\r\n\r\nThanks,\r\nC\r\n ",
    "84263": "@Halla Yang Could you tell me, what size you use?\r\n",
    "84358": "Hey guys! Could somebody please kindly explain me, why I got down votes to my last question, while more other ones share their thoughts regarding image size? Thank you in advance.",
    "84253": "thank you!\r\n\r\nlooks it's too much for my laptop.\r\n",
    "84232": ""
  }
}