{
  "id": 156696,
  "title": "Shouldn't FashionODFinal.ipynb take in 224x224, but the dataset are not fixed to this shape?!",
  "url": "/competitions/til2020/discussion/156696",
  "author_name": "",
  "post_date": "2020-06-07T11:41:41.880476400Z",
  "votes": 2,
  "comment_count": 2,
  "views": 0,
  "content": "<p>There are 2 versions of the dataset:\n1. From TILPickle, which gives the data in 224x224 images\n2. From TILSequence, which gives the flexibility to experiment with different image sizes, but need to adapt the model to process different sizes</p>\n\n<p>Look deeper into the code cells within FashionODFinal.ipynb on how to create TILPickle and TILSequence! </p>",
  "messages": [
    {
      "id": "877172",
      "postDate": "06/07/2020 11:41:41",
      "content": "<p>There are 2 versions of the dataset:\n1. From TILPickle, which gives the data in 224x224 images\n2. From TILSequence, which gives the flexibility to experiment with different image sizes, but need to adapt the model to process different sizes</p>\n\n<p>Look deeper into the code cells within FashionODFinal.ipynb on how to create TILPickle and TILSequence! </p>",
      "rawMarkdown": "There are 2 versions of the dataset:\n1. From TILPickle, which gives the data in 224x224 images\n2. From TILSequence, which gives the flexibility to experiment with different image sizes, but need to adapt the model to process different sizes\n\nLook deeper into the code cells within FashionODFinal.ipynb on how to create TILPickle and TILSequence!",
      "votes": null
    },
    {
      "id": "881688",
      "postDate": "06/11/2020 09:06:32",
      "content": "<p>Hi for the val.json file and train.json files, I presume they are for usage in the NLP component?</p>\n\n<p>Also when I tried to read_json the val.json file, there was an error saying that \"ValueError: arrays must all be the same length\". Is there any recommendation regarding this?</p>",
      "rawMarkdown": "Hi for the val.json file and train.json files, I presume they are for usage in the NLP component?\n\nAlso when I tried to read_json the val.json file, there was an error saying that \"ValueError: arrays must all be the same length\". Is there any recommendation regarding this?",
      "votes": null
    },
    {
      "id": "881863",
      "postDate": "06/11/2020 12:25:33",
      "content": "<p>No, they're for the CV component; they contain the bounding box descriptions </p>",
      "rawMarkdown": "No, they're for the CV component; they contain the bounding box descriptions",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 881688,
      "author_name": "quaternionande",
      "author_url": "",
      "post_date": "06/11/2020 09:06:32",
      "content": "<p>Hi for the val.json file and train.json files, I presume they are for usage in the NLP component?</p>\n\n<p>Also when I tried to read_json the val.json file, there was an error saying that \"ValueError: arrays must all be the same length\". Is there any recommendation regarding this?</p>",
      "votes": null,
      "replies": [
        {
          "id": 881863,
          "author_name": "htet96",
          "author_url": "",
          "post_date": "06/11/2020 12:25:33",
          "content": "<p>No, they're for the CV component; they contain the bounding box descriptions </p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "877172": "There are 2 versions of the dataset:\n1. From TILPickle, which gives the data in 224x224 images\n2. From TILSequence, which gives the flexibility to experiment with different image sizes, but need to adapt the model to process different sizes\n\nLook deeper into the code cells within FashionODFinal.ipynb on how to create TILPickle and TILSequence!",
    "881688": "Hi for the val.json file and train.json files, I presume they are for usage in the NLP component?\n\nAlso when I tried to read_json the val.json file, there was an error saying that \"ValueError: arrays must all be the same length\". Is there any recommendation regarding this?",
    "881863": "No, they're for the CV component; they contain the bounding box descriptions"
  },
  "source": "meta"
}