{
  "id": 173247,
  "title": "Anyone with tfrecords of Training Dataset ?",
  "url": "/competitions/landmark-retrieval-2020/discussion/173247",
  "author_name": "",
  "post_date": "2020-08-08T13:09:43.591388200Z",
  "votes": 1,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Hi everyone. I am following the training pipeline from this link to beat the baseline.  It creates the tfrecords of training and the validation datasets which take about 12 hours and 500 GB of space. I am just curious has anyone tried it and created the tfrecords ? If yes. Are they willing to share them cause this will make the competition more accessible for people.Thanks </p>\n<p><a href=\"https://github.com/tensorflow/models/blob/master/research/delf/delf/python/training/README.md\" target=\"_blank\">https://github.com/tensorflow/models/blob/master/research/delf/delf/python/training/README.md</a></p>",
  "messages": [
    {
      "id": "962821",
      "postDate": "08/08/2020 13:09:43",
      "content": "<p>Hi everyone. I am following the training pipeline from this link to beat the baseline.  It creates the tfrecords of training and the validation datasets which take about 12 hours and 500 GB of space. I am just curious has anyone tried it and created the tfrecords ? If yes. Are they willing to share them cause this will make the competition more accessible for people.Thanks </p>\n<p><a href=\"https://github.com/tensorflow/models/blob/master/research/delf/delf/python/training/README.md\" target=\"_blank\">https://github.com/tensorflow/models/blob/master/research/delf/delf/python/training/README.md</a></p>",
      "rawMarkdown": "Hi everyone. I am following the training pipeline from this link to beat the baseline.  It creates the tfrecords of training and the validation datasets which take about 12 hours and 500 GB of space. I am just curious has anyone tried it and created the tfrecords ? If yes. Are they willing to share them cause this will make the competition more accessible for people.Thanks \n\n\nhttps://github.com/tensorflow/models/blob/master/research/delf/delf/python/training/README.md",
      "votes": null
    },
    {
      "id": "963037",
      "postDate": "08/08/2020 16:07:32",
      "content": "<p>Hope this helps...\n<a href=\"https://www.kaggle.com/c/landmark-retrieval-2020/discussion/171297\">https://www.kaggle.com/c/landmark-retrieval-2020/discussion/171297</a></p>",
      "rawMarkdown": "Hope this helps...\nhttps://www.kaggle.com/c/landmark-retrieval-2020/discussion/171297",
      "votes": null
    },
    {
      "id": "963246",
      "postDate": "08/08/2020 19:39:09",
      "content": "<p>Hi. Thank you for your amazing efforts. But this just contains the Validation TFRecords. I need the training one too. I can generate the records but due to limitation of resources cannot at the moment. I have read the paper of DELG. Just needed the TFRecords of  whole dataset to nail in the competition. Is it possible to share some Google Cloud Storage link ?</p>",
      "rawMarkdown": "Hi. Thank you for your amazing efforts. But this just contains the Validation TFRecords. I need the training one too. I can generate the records but due to limitation of resources cannot at the moment. I have read the paper of DELG. Just needed the TFRecords of  whole dataset to nail in the competition. Is it possible to share some Google Cloud Storage link ?",
      "votes": null
    },
    {
      "id": "963270",
      "postDate": "08/08/2020 20:19:19",
      "content": "<p>Hey, I don't have much cloud storage available. And the train tf records are worth ~100 GB. Let me see what I can do. Give me a day for that.</p>",
      "rawMarkdown": "Hey, I don't have much cloud storage available. And the train tf records are worth ~100 GB. Let me see what I can do. Give me a day for that.",
      "votes": null
    },
    {
      "id": "964406",
      "postDate": "08/09/2020 20:27:14",
      "content": "<p><a href=\"https://www.kaggle.com/c/landmark-retrieval-2020/discussion/173572\">https://www.kaggle.com/c/landmark-retrieval-2020/discussion/173572</a></p>",
      "rawMarkdown": "https://www.kaggle.com/c/landmark-retrieval-2020/discussion/173572",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 963037,
      "author_name": "chandanverma",
      "author_url": "",
      "post_date": "08/08/2020 16:07:32",
      "content": "<p>Hope this helps...\n<a href=\"https://www.kaggle.com/c/landmark-retrieval-2020/discussion/171297\">https://www.kaggle.com/c/landmark-retrieval-2020/discussion/171297</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 963246,
          "author_name": "jamshaidsohail5",
          "author_url": "",
          "post_date": "08/08/2020 19:39:09",
          "content": "<p>Hi. Thank you for your amazing efforts. But this just contains the Validation TFRecords. I need the training one too. I can generate the records but due to limitation of resources cannot at the moment. I have read the paper of DELG. Just needed the TFRecords of  whole dataset to nail in the competition. Is it possible to share some Google Cloud Storage link ?</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 963270,
          "author_name": "chandanverma",
          "author_url": "",
          "post_date": "08/08/2020 20:19:19",
          "content": "<p>Hey, I don't have much cloud storage available. And the train tf records are worth ~100 GB. Let me see what I can do. Give me a day for that.</p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 964406,
          "author_name": "chandanverma",
          "author_url": "",
          "post_date": "08/09/2020 20:27:14",
          "content": "<p><a href=\"https://www.kaggle.com/c/landmark-retrieval-2020/discussion/173572\">https://www.kaggle.com/c/landmark-retrieval-2020/discussion/173572</a></p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "962821": "Hi everyone. I am following the training pipeline from this link to beat the baseline.  It creates the tfrecords of training and the validation datasets which take about 12 hours and 500 GB of space. I am just curious has anyone tried it and created the tfrecords ? If yes. Are they willing to share them cause this will make the competition more accessible for people.Thanks \n\n\nhttps://github.com/tensorflow/models/blob/master/research/delf/delf/python/training/README.md",
    "963037": "Hope this helps...\nhttps://www.kaggle.com/c/landmark-retrieval-2020/discussion/171297",
    "963246": "Hi. Thank you for your amazing efforts. But this just contains the Validation TFRecords. I need the training one too. I can generate the records but due to limitation of resources cannot at the moment. I have read the paper of DELG. Just needed the TFRecords of  whole dataset to nail in the competition. Is it possible to share some Google Cloud Storage link ?",
    "963270": "Hey, I don't have much cloud storage available. And the train tf records are worth ~100 GB. Let me see what I can do. Give me a day for that.",
    "964406": "https://www.kaggle.com/c/landmark-retrieval-2020/discussion/173572"
  },
  "source": "meta"
}