{
  "id": 60573,
  "title": "Are image ids unique across training and validation sets?",
  "url": "/competitions/google-ai-open-images-object-detection-track/discussion/60573",
  "author_name": "",
  "post_date": "2018-07-06T20:18:14.235386800Z",
  "votes": 1,
  "comment_count": 1,
  "views": 0,
  "content": "<p>I am going to use a pretrained YOLOv3 implementation by andy-yun(available on github). The images for training AND validation sets are stored in the same folder. The Dataset class will read the images using 2 lists of paths in a txt. In order for there not be any conflicts, IDs for images must be unique across both training and validation. Does anyone have any prior knowledge on this?</p>",
  "messages": [
    {
      "id": "353472",
      "postDate": "07/06/2018 20:18:14",
      "content": "<p>I am going to use a pretrained YOLOv3 implementation by andy-yun(available on github). The images for training AND validation sets are stored in the same folder. The Dataset class will read the images using 2 lists of paths in a txt. In order for there not be any conflicts, IDs for images must be unique across both training and validation. Does anyone have any prior knowledge on this?</p>",
      "rawMarkdown": "I am going to use a pretrained YOLOv3 implementation by andy-yun(available on github). The images for training AND validation sets are stored in the same folder. The Dataset class will read the images using 2 lists of paths in a txt. In order for there not be any conflicts, IDs for images must be unique across both training and validation. Does anyone have any prior knowledge on this?",
      "votes": null
    },
    {
      "id": "354313",
      "postDate": "07/09/2018 09:53:34",
      "content": "<p>The validation image ID list given is a subset of the train image list. After you separate, the image ID's will be unique.</p>",
      "rawMarkdown": "The validation image ID list given is a subset of the train image list. After you separate, the image ID's will be unique.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 354313,
      "author_name": "rupraze",
      "author_url": "",
      "post_date": "07/09/2018 09:53:34",
      "content": "<p>The validation image ID list given is a subset of the train image list. After you separate, the image ID's will be unique.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "353472": "I am going to use a pretrained YOLOv3 implementation by andy-yun(available on github). The images for training AND validation sets are stored in the same folder. The Dataset class will read the images using 2 lists of paths in a txt. In order for there not be any conflicts, IDs for images must be unique across both training and validation. Does anyone have any prior knowledge on this?",
    "354313": "The validation image ID list given is a subset of the train image list. After you separate, the image ID's will be unique."
  },
  "source": "meta"
}