{
  "id": 108668,
  "title": "Error in map.json 'filename'",
  "url": "/competitions/3d-object-detection-for-autonomous-vehicles/discussion/108668",
  "author_name": "",
  "post_date": "2019-09-13T07:20:28.642633Z",
  "votes": null,
  "comment_count": 3,
  "views": 0,
  "content": "<p>I'm trying to get the <strong>tutorial_lyft.ipynb</strong> in <strong>nuscenes-devkit</strong> to work, and was stuck at the dataset loading after I modified the file paths to Kaggle's dataset structure.</p>\n\n<p>```</p>\n\n<h1>Load the SDK</h1>\n\n<p>%matplotlib inline\nfrom lyft_dataset_sdk.lyftdataset import LyftDataset</p>\n\n<h1>Load the dataset</h1>\n\n<h1>Adjust the dataroot parameter below to point to your local dataset path.</h1>\n\n<h1>The correct dataset path contains at least the following four folders (or similar): images, lidar, maps, v1.0.1-train</h1>\n\n<p>level5data = LyftDataset(data_path='../input/3d-object-detection-for-autonomous-vehicles', \n                         json_path='../input/3d-object-detection-for-autonomous-vehicles/train_data/', verbose=True)\n```\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1822666%2Fa424e92448fefb238d31b7f61817bb4a%2Flyft%20error.PNG?generation=1568359020681587&amp;alt=media\" alt=\"\"></p>\n\n<p>After digging into the <strong>map.json</strong> for <code>train_data</code> and <code>test_data</code>, I found that the <code>filename</code> are pointing to the incorrect folders:</p>\n\n<p><code>'filename': 'maps/map_raster_palo_alto.png'</code> </p>\n\n<p>The correct paths should points to the train &amp; test folders respectively:</p>\n\n<p><code>'filename': 'train_maps/map_raster_palo_alto.png'</code> for <code>train_data/map.json</code>\n<code>'filename': 'test_maps/map_raster_palo_alto.png'</code>  for <code>test_data/map.json</code></p>\n\n<p>Could you please update that so that we can at least make use of the tutorial notebook easily?</p>\n\n<p>Thanks.</p>",
  "messages": [
    {
      "id": "625547",
      "postDate": "09/13/2019 07:20:28",
      "content": "<p>I'm trying to get the <strong>tutorial_lyft.ipynb</strong> in <strong>nuscenes-devkit</strong> to work, and was stuck at the dataset loading after I modified the file paths to Kaggle's dataset structure.</p>\n\n<p>```</p>\n\n<h1>Load the SDK</h1>\n\n<p>%matplotlib inline\nfrom lyft_dataset_sdk.lyftdataset import LyftDataset</p>\n\n<h1>Load the dataset</h1>\n\n<h1>Adjust the dataroot parameter below to point to your local dataset path.</h1>\n\n<h1>The correct dataset path contains at least the following four folders (or similar): images, lidar, maps, v1.0.1-train</h1>\n\n<p>level5data = LyftDataset(data_path='../input/3d-object-detection-for-autonomous-vehicles', \n                         json_path='../input/3d-object-detection-for-autonomous-vehicles/train_data/', verbose=True)\n```\n<img src=\"https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1822666%2Fa424e92448fefb238d31b7f61817bb4a%2Flyft%20error.PNG?generation=1568359020681587&amp;alt=media\" alt=\"\"></p>\n\n<p>After digging into the <strong>map.json</strong> for <code>train_data</code> and <code>test_data</code>, I found that the <code>filename</code> are pointing to the incorrect folders:</p>\n\n<p><code>'filename': 'maps/map_raster_palo_alto.png'</code> </p>\n\n<p>The correct paths should points to the train &amp; test folders respectively:</p>\n\n<p><code>'filename': 'train_maps/map_raster_palo_alto.png'</code> for <code>train_data/map.json</code>\n<code>'filename': 'test_maps/map_raster_palo_alto.png'</code>  for <code>test_data/map.json</code></p>\n\n<p>Could you please update that so that we can at least make use of the tutorial notebook easily?</p>\n\n<p>Thanks.</p>",
      "rawMarkdown": "I'm trying to get the **tutorial_lyft.ipynb** in **nuscenes-devkit** to work, and was stuck at the dataset loading after I modified the file paths to Kaggle's dataset structure.\n\n```\n# Load the SDK\n%matplotlib inline\nfrom lyft_dataset_sdk.lyftdataset import LyftDataset\n\n# Load the dataset\n# Adjust the dataroot parameter below to point to your local dataset path.\n# The correct dataset path contains at least the following four folders (or similar): images, lidar, maps, v1.0.1-train\nlevel5data = LyftDataset(data_path='../input/3d-object-detection-for-autonomous-vehicles', \n                         json_path='../input/3d-object-detection-for-autonomous-vehicles/train_data/', verbose=True)\n```\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1822666%2Fa424e92448fefb238d31b7f61817bb4a%2Flyft%20error.PNG?generation=1568359020681587&amp;alt=media)\n\n\nAfter digging into the **map.json** for `train_data` and `test_data`, I found that the `filename` are pointing to the incorrect folders:\n\n`'filename': 'maps/map_raster_palo_alto.png'` \n\nThe correct paths should points to the train &amp; test folders respectively:\n\n`'filename': 'train_maps/map_raster_palo_alto.png'` for `train_data/map.json`\n`'filename': 'test_maps/map_raster_palo_alto.png'`  for `test_data/map.json`\n\nCould you please update that so that we can at least make use of the tutorial notebook easily?\n\nThanks.",
      "votes": null
    },
    {
      "id": "625567",
      "postDate": "09/13/2019 07:32:41",
      "content": "<p><a href=\"https://www.kaggle.com/seshurajup/lyft-level-5-av-dataset-notebook-from-github#625566\">https://www.kaggle.com/seshurajup/lyft-level-5-av-dataset-notebook-from-github#625566</a></p>",
      "rawMarkdown": "https://www.kaggle.com/seshurajup/lyft-level-5-av-dataset-notebook-from-github#625566",
      "votes": null
    },
    {
      "id": "625580",
      "postDate": "09/13/2019 07:39:43",
      "content": "<p>your approach works, thanks.</p>",
      "rawMarkdown": "your approach works, thanks.",
      "votes": null
    },
    {
      "id": "636076",
      "postDate": "09/28/2019 19:21:01",
      "content": "<p>What I did was simply put symbolic links in the directory with the names that the devkit was looking for.  For instance:\n<code>ln -s train_maps maps</code>\n<code>ln -s train_images images</code>\n<code>ln -s train_data data</code>\n<code>ln -s train_lidar lidar</code>\nAnd that did the trick for me.  If you are not on a linux system, maybe try the equivalent on whatever operating system you're using, and I think that would work.  You could also swap the links to point to test_*, which will allow you to explore the test dataset.\n-Dan McGonigle</p>",
      "rawMarkdown": "What I did was simply put symbolic links in the directory with the names that the devkit was looking for.  For instance:\n`ln -s train_maps maps`\n`ln -s train_images images`\n`ln -s train_data data`\n`ln -s train_lidar lidar`\nAnd that did the trick for me.  If you are not on a linux system, maybe try the equivalent on whatever operating system you're using, and I think that would work.  You could also swap the links to point to test_*, which will allow you to explore the test dataset.\n-Dan McGonigle",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 625567,
      "author_name": "rishabhiitbhu",
      "author_url": "",
      "post_date": "09/13/2019 07:32:41",
      "content": "<p><a href=\"https://www.kaggle.com/seshurajup/lyft-level-5-av-dataset-notebook-from-github#625566\">https://www.kaggle.com/seshurajup/lyft-level-5-av-dataset-notebook-from-github#625566</a></p>",
      "votes": null,
      "replies": [
        {
          "id": 625580,
          "author_name": "chewzy",
          "author_url": "",
          "post_date": "09/13/2019 07:39:43",
          "content": "<p>your approach works, thanks.</p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 636076,
      "author_name": "dpmcgonigle",
      "author_url": "",
      "post_date": "09/28/2019 19:21:01",
      "content": "<p>What I did was simply put symbolic links in the directory with the names that the devkit was looking for.  For instance:\n<code>ln -s train_maps maps</code>\n<code>ln -s train_images images</code>\n<code>ln -s train_data data</code>\n<code>ln -s train_lidar lidar</code>\nAnd that did the trick for me.  If you are not on a linux system, maybe try the equivalent on whatever operating system you're using, and I think that would work.  You could also swap the links to point to test_*, which will allow you to explore the test dataset.\n-Dan McGonigle</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "625547": "I'm trying to get the **tutorial_lyft.ipynb** in **nuscenes-devkit** to work, and was stuck at the dataset loading after I modified the file paths to Kaggle's dataset structure.\n\n```\n# Load the SDK\n%matplotlib inline\nfrom lyft_dataset_sdk.lyftdataset import LyftDataset\n\n# Load the dataset\n# Adjust the dataroot parameter below to point to your local dataset path.\n# The correct dataset path contains at least the following four folders (or similar): images, lidar, maps, v1.0.1-train\nlevel5data = LyftDataset(data_path='../input/3d-object-detection-for-autonomous-vehicles', \n                         json_path='../input/3d-object-detection-for-autonomous-vehicles/train_data/', verbose=True)\n```\n![](https://www.googleapis.com/download/storage/v1/b/kaggle-user-content/o/inbox%2F1822666%2Fa424e92448fefb238d31b7f61817bb4a%2Flyft%20error.PNG?generation=1568359020681587&amp;alt=media)\n\n\nAfter digging into the **map.json** for `train_data` and `test_data`, I found that the `filename` are pointing to the incorrect folders:\n\n`'filename': 'maps/map_raster_palo_alto.png'` \n\nThe correct paths should points to the train &amp; test folders respectively:\n\n`'filename': 'train_maps/map_raster_palo_alto.png'` for `train_data/map.json`\n`'filename': 'test_maps/map_raster_palo_alto.png'`  for `test_data/map.json`\n\nCould you please update that so that we can at least make use of the tutorial notebook easily?\n\nThanks.",
    "625567": "https://www.kaggle.com/seshurajup/lyft-level-5-av-dataset-notebook-from-github#625566",
    "625580": "your approach works, thanks.",
    "636076": "What I did was simply put symbolic links in the directory with the names that the devkit was looking for.  For instance:\n`ln -s train_maps maps`\n`ln -s train_images images`\n`ln -s train_data data`\n`ln -s train_lidar lidar`\nAnd that did the trick for me.  If you are not on a linux system, maybe try the equivalent on whatever operating system you're using, and I think that would work.  You could also swap the links to point to test_*, which will allow you to explore the test dataset.\n-Dan McGonigle"
  },
  "source": "meta"
}