{
  "id": 113945,
  "title": "External dataset for car detection?",
  "url": "/competitions/pku-autonomous-driving/discussion/113945",
  "author_name": "",
  "post_date": "2019-10-23T07:24:24.599867700Z",
  "votes": 5,
  "comment_count": 4,
  "views": 0,
  "content": "<p>Since the training set only has 4000+ images, the performance of detection may not be so good. Can we found external datasets for the pre-training?</p>\n\n<p>For starter, https://www.kaggle.com/c/carvana-image-masking-challenge is  a car segmentation competition, where compare with this one, the scene in Carvana is much easier. Therefore, I'm not sure whether it is useful.\nIn addition, since the 3D model is provided, or we can produce synthetic images via stick the models into flick images?</p>",
  "messages": [
    {
      "id": "655560",
      "postDate": "10/23/2019 07:24:24",
      "content": "<p>Since the training set only has 4000+ images, the performance of detection may not be so good. Can we found external datasets for the pre-training?</p>\n\n<p>For starter, https://www.kaggle.com/c/carvana-image-masking-challenge is  a car segmentation competition, where compare with this one, the scene in Carvana is much easier. Therefore, I'm not sure whether it is useful.\nIn addition, since the 3D model is provided, or we can produce synthetic images via stick the models into flick images?</p>",
      "rawMarkdown": "Since the training set only has 4000+ images, the performance of detection may not be so good. Can we found external datasets for the pre-training?\n\nFor starter, [https://www.kaggle.com/c/carvana-image-masking-challenge](Carvana) is  a car segmentation competition, where compare with this one, the scene in Carvana is much easier. Therefore, I'm not sure whether it is useful.\nIn addition, since the 3D model is provided, or we can produce synthetic images via stick the models into flick images?",
      "votes": null
    },
    {
      "id": "655574",
      "postDate": "10/23/2019 07:54:35",
      "content": "<p>Cannot open \".pkl\" models, both via python2 and python3.</p>\n\n<p>For py3, when I installed \"objloader\", I got this error:\n<em>AttributeError: Can't get attribute 'CHJtinyobj' on</em>\nFor py2, the error is:\n<em>Exception: Python 2 is not supported!</em></p>",
      "rawMarkdown": "Cannot open \".pkl\" models, both via python2 and python3.\n\nFor py3, when I installed \"objloader\", I got this error:\n*AttributeError: Can't get attribute 'CHJtinyobj' on*\nFor py2, the error is:\n*Exception: Python 2 is not supported!*",
      "votes": null
    },
    {
      "id": "656384",
      "postDate": "10/24/2019 07:28:12",
      "content": "<p>Well, seems json files are uploaded by organizers : ) <a href=\"https://www.kaggle.com/c/pku-autonomous-driving/download/car_models_json.zip\">https://www.kaggle.com/c/pku-autonomous-driving/download/car_models_json.zip</a></p>",
      "rawMarkdown": "Well, seems json files are uploaded by organizers : ) https://www.kaggle.com/c/pku-autonomous-driving/download/car_models_json.zip",
      "votes": null
    },
    {
      "id": "656427",
      "postDate": "10/24/2019 09:06:13",
      "content": "<p>James found some datasets, \n<a href=\"https://www.kaggle.com/c/pku-autonomous-driving/discussion/113918#latest-655432\">https://www.kaggle.com/c/pku-autonomous-driving/discussion/113918#latest-655432</a></p>",
      "rawMarkdown": "James found some datasets, \nhttps://www.kaggle.com/c/pku-autonomous-driving/discussion/113918#latest-655432",
      "votes": null
    },
    {
      "id": "656727",
      "postDate": "10/24/2019 15:06:00",
      "content": "<p>I met the same error as you. </p>",
      "rawMarkdown": "I met the same error as you.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 655574,
      "author_name": "yiheng",
      "author_url": "",
      "post_date": "10/23/2019 07:54:35",
      "content": "<p>Cannot open \".pkl\" models, both via python2 and python3.</p>\n\n<p>For py3, when I installed \"objloader\", I got this error:\n<em>AttributeError: Can't get attribute 'CHJtinyobj' on</em>\nFor py2, the error is:\n<em>Exception: Python 2 is not supported!</em></p>",
      "votes": null,
      "replies": [
        {
          "id": 656384,
          "author_name": "yiheng",
          "author_url": "",
          "post_date": "10/24/2019 07:28:12",
          "content": "<p>Well, seems json files are uploaded by organizers : ) <a href=\"https://www.kaggle.com/c/pku-autonomous-driving/download/car_models_json.zip\">https://www.kaggle.com/c/pku-autonomous-driving/download/car_models_json.zip</a></p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 656727,
          "author_name": "ericji",
          "author_url": "",
          "post_date": "10/24/2019 15:06:00",
          "content": "<p>I met the same error as you. </p>",
          "votes": null,
          "replies": []
        }
      ]
    },
    {
      "id": 656427,
      "author_name": "yiheng",
      "author_url": "",
      "post_date": "10/24/2019 09:06:13",
      "content": "<p>James found some datasets, \n<a href=\"https://www.kaggle.com/c/pku-autonomous-driving/discussion/113918#latest-655432\">https://www.kaggle.com/c/pku-autonomous-driving/discussion/113918#latest-655432</a></p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "655560": "Since the training set only has 4000+ images, the performance of detection may not be so good. Can we found external datasets for the pre-training?\n\nFor starter, [https://www.kaggle.com/c/carvana-image-masking-challenge](Carvana) is  a car segmentation competition, where compare with this one, the scene in Carvana is much easier. Therefore, I'm not sure whether it is useful.\nIn addition, since the 3D model is provided, or we can produce synthetic images via stick the models into flick images?",
    "655574": "Cannot open \".pkl\" models, both via python2 and python3.\n\nFor py3, when I installed \"objloader\", I got this error:\n*AttributeError: Can't get attribute 'CHJtinyobj' on*\nFor py2, the error is:\n*Exception: Python 2 is not supported!*",
    "656384": "Well, seems json files are uploaded by organizers : ) https://www.kaggle.com/c/pku-autonomous-driving/download/car_models_json.zip",
    "656427": "James found some datasets, \nhttps://www.kaggle.com/c/pku-autonomous-driving/discussion/113918#latest-655432",
    "656727": "I met the same error as you."
  },
  "source": "meta"
}