{
  "id": 70680,
  "title": "draw time and pause time",
  "url": "/competitions/quickdraw-doodle-recognition/discussion/70680",
  "author_name": "hengck23",
  "post_date": "2018-11-06T14:10:24.840000",
  "votes": 9,
  "comment_count": 12,
  "views": 0,
  "content": "<p>I read that quick draw dataset has \"draw time\" and\" pause time\" attribute. Is it true? </p>\n\n<p><img src=\"https://i1.wp.com/teaching.statistics-is-awesome.org/wp-content/uploads/data-card.png\" alt=\"enter image description here\"></p>",
  "messages": [
    {
      "id": 416284,
      "postDate": "2018-11-06T14:10:24.840Z",
      "content": "<p>I read that quick draw dataset has \"draw time\" and\" pause time\" attribute. Is it true? </p>\n\n<p><img src=\"https://i1.wp.com/teaching.statistics-is-awesome.org/wp-content/uploads/data-card.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "I read that quick draw dataset has \"draw time\" and\" pause time\" attribute. Is it true? \n\n![enter image description here][1]\n\n\n  [1]: https://i1.wp.com/teaching.statistics-is-awesome.org/wp-content/uploads/data-card.png",
      "votes": 9
    },
    {
      "id": 418298,
      "postDate": "2018-11-09T15:59:40.597Z",
      "content": "<p>using time to encode attention:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418298/10650/encode_time.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "using time to encode attention:\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/418298/10650/encode_time.png",
      "votes": 1
    },
    {
      "id": 416384,
      "postDate": "2018-11-06T15:54:12.817Z",
      "content": "<p>ok, i found it:</p>\n\n<p><a href=\"https://github.com/vlandham/quickdraw/blob/master/analysis/all_aggregates.csv\">https://github.com/vlandham/quickdraw/blob/master/analysis/all_aggregates.csv</a></p>\n\n<p><a href=\"https://github.com/vlandham/quickdraw/blob/master/analysis/extract_times.ipynb\">https://github.com/vlandham/quickdraw/blob/master/analysis/extract_times.ipynb</a></p>\n\n<pre><code>def first_time(stroke):\n     return stroke[2][0]\n</code></pre>\n\n<p>...</p>\n\n<p>def gen_stats_stroke(drawing, stroke, stroke_index):</p>\n\n<pre><code>last_stroke = drawing['drawing'][len(drawing['drawing']) - 1]\nstats = {\n    #\"key_id\": drawing['key_id'],\n    #\"recognized\": drawing['recognized'],\n    #\"word\": drawing['word'],\n    #\"countrycode\": drawing['countrycode'],\n    #\"drawing_time\": last_time(last_stroke),\n    #\"stroke_count\": len(drawing['drawing']),\n    \"stroke_index\": stroke_index,\n    \"stroke_time\": total_time(stroke),\n    \"stroke_time_first\": first_time(stroke),\n    \"stroke_time_last\": last_time(stroke),\n    \"stroke_time_pause\": pause_before(drawing, stroke, stroke_index),\n    \"stroke_time_min\": min_time(stroke),\n    \"stroke_time_max\": max_time(stroke),\n    \"stroke_time_mixedup\": mixed_up_time(stroke)\n}\n\nreturn stats\n</code></pre>",
      "rawMarkdown": "ok, i found it:\n\nhttps://github.com/vlandham/quickdraw/blob/master/analysis/all_aggregates.csv\n\nhttps://github.com/vlandham/quickdraw/blob/master/analysis/extract_times.ipynb\n\n\n\n\n    def first_time(stroke):\n         return stroke[2][0]\n\n...\n\n   def gen_stats_stroke(drawing, stroke, stroke_index):\n    \n    last_stroke = drawing['drawing'][len(drawing['drawing']) - 1]\n    stats = {\n        #\"key_id\": drawing['key_id'],\n        #\"recognized\": drawing['recognized'],\n        #\"word\": drawing['word'],\n        #\"countrycode\": drawing['countrycode'],\n        #\"drawing_time\": last_time(last_stroke),\n        #\"stroke_count\": len(drawing['drawing']),\n        \"stroke_index\": stroke_index,\n        \"stroke_time\": total_time(stroke),\n        \"stroke_time_first\": first_time(stroke),\n        \"stroke_time_last\": last_time(stroke),\n        \"stroke_time_pause\": pause_before(drawing, stroke, stroke_index),\n        \"stroke_time_min\": min_time(stroke),\n        \"stroke_time_max\": max_time(stroke),\n        \"stroke_time_mixedup\": mixed_up_time(stroke)\n    }\n    \n    return stats",
      "votes": 1,
      "replies": [
        {
          "id": 420206,
          "postDate": "2018-11-13T09:07:08.923Z",
          "content": "<p>Hi Heng. Does the dataset you mentioned here in github the same as kaggle provide? Thus we have to download it for the stroke time information right?</p>",
          "rawMarkdown": "Hi Heng. Does the dataset you mentioned here in github the same as kaggle provide? Thus we have to download it for the stroke time information right?"
        }
      ]
    },
    {
      "id": 425796,
      "postDate": "2018-11-22T05:51:35.180Z",
      "content": "<p>I can't found this info for test data. Only country code</p>",
      "rawMarkdown": "I can't found this info for test data. Only country code",
      "replies": [
        {
          "id": 427051,
          "postDate": "2018-11-24T12:44:08.350Z",
          "content": "<p>It exists only in raw data, third item in each stroke array.</p>",
          "rawMarkdown": "It exists only in raw data, third item in each stroke array.",
          "votes": 1
        }
      ]
    },
    {
      "id": 420250,
      "postDate": "2018-11-13T10:51:16.803Z",
      "content": "<p>encode using draw and pause time.</p>\n\n<p>note that the \"person\" become \"evident\". </p>\n\n<p>hence the correct label is likely to be a bed</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/420250/10668/0.40_9591733508469962_cooler_oven_diving_board.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "encode using draw and pause time.\n\nnote that the \"person\" become \"evident\". \n\nhence the correct label is likely to be a bed\n\n\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/420250/10668/0.40_9591733508469962_cooler_oven_diving_board.png"
    },
    {
      "id": 419281,
      "postDate": "2018-11-11T16:21:30.477Z",
      "content": "<p>using strokes for part attention:</p>\n\n<p>\"Weakly Supervised Local Attention Network for Fine-Grained Visual Classification\"\n- Tao Hu, Arvix 2018</p>\n\n<p>can also augment auxiliary features (e.g. stroke length, time, etc) to each part feature</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/419281/10656/attention_pooling.png\" alt=\"enter image description here\"></p>\n\n<hr>\n\n<p>related:</p>\n\n<p>interacting part features:</p>\n\n<p>\"Low-rank Bilinear Pooling for Fine-Grained Classification\"- Shu Kong, Arvix 2016</p>",
      "rawMarkdown": "using strokes for part attention:\n\n\"Weakly Supervised Local Attention Network for Fine-Grained Visual Classification\"\n- Tao Hu, Arvix 2018\n\ncan also augment auxiliary features (e.g. stroke length, time, etc) to each part feature\n\n   ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/419281/10656/attention_pooling.png\n\n\n---\n\nrelated:\n\ninteracting part features:\n\n\"Low-rank Bilinear Pooling for Fine-Grained Classification\"- Shu Kong, Arvix 2016"
    },
    {
      "id": 418310,
      "postDate": "2018-11-09T16:19:03.683Z",
      "content": "<p>attention on the bridge:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10652/bridge1.png\" alt=\"enter image description here\">\n   <img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10653/bridge2.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "attention on the bridge:\n\n\n   ![enter image description here][1]\n   ![enter image description here][2]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10652/bridge1.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10653/bridge2.png"
    },
    {
      "id": 418302,
      "postDate": "2018-11-09T16:07:08.883Z",
      "content": "<p>here are more attention map:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418302/10651/encode_attent2.png\" alt=\"enter image description here\"></p>",
      "rawMarkdown": "here are more attention map:\n\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/418302/10651/encode_attent2.png"
    },
    {
      "id": 418284,
      "postDate": "2018-11-09T15:36:01.527Z",
      "content": "<p>check the raw data!</p>\n\n<p>the information is different. seems that raw data has timestamp per point?</p>",
      "rawMarkdown": "check the raw data!\n\nthe information is different. seems that raw data has timestamp per point?",
      "replies": [
        {
          "id": 418321,
          "postDate": "2018-11-09T16:37:37.060Z",
          "content": "<p>I think so. There is no one-on-one mapping between simplified and raw points, but there is one-on-one mapping for strokes. So if we choose to work on simplified CSVs, we can only use the time information stroke-wise. </p>",
          "rawMarkdown": "I think so. There is no one-on-one mapping between simplified and raw points, but there is one-on-one mapping for strokes. So if we choose to work on simplified CSVs, we can only use the time information stroke-wise. "
        },
        {
          "id": 418961,
          "postDate": "2018-11-11T01:04:40.097Z",
          "content": "<p>Yes, the dataset description page on GitHab says so. </p>",
          "rawMarkdown": "Yes, the dataset description page on GitHab says so. "
        }
      ]
    }
  ],
  "comments": [
    {
      "id": 418298,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-09T15:59:40.597000",
      "content": "<p>using time to encode attention:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418298/10650/encode_time.png\" alt=\"enter image description here\"></p>",
      "votes": 1,
      "replies": []
    },
    {
      "id": 416384,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-06T15:54:12.817000",
      "content": "<p>ok, i found it:</p>\n\n<p><a href=\"https://github.com/vlandham/quickdraw/blob/master/analysis/all_aggregates.csv\">https://github.com/vlandham/quickdraw/blob/master/analysis/all_aggregates.csv</a></p>\n\n<p><a href=\"https://github.com/vlandham/quickdraw/blob/master/analysis/extract_times.ipynb\">https://github.com/vlandham/quickdraw/blob/master/analysis/extract_times.ipynb</a></p>\n\n<pre><code>def first_time(stroke):\n     return stroke[2][0]\n</code></pre>\n\n<p>...</p>\n\n<p>def gen_stats_stroke(drawing, stroke, stroke_index):</p>\n\n<pre><code>last_stroke = drawing['drawing'][len(drawing['drawing']) - 1]\nstats = {\n    #\"key_id\": drawing['key_id'],\n    #\"recognized\": drawing['recognized'],\n    #\"word\": drawing['word'],\n    #\"countrycode\": drawing['countrycode'],\n    #\"drawing_time\": last_time(last_stroke),\n    #\"stroke_count\": len(drawing['drawing']),\n    \"stroke_index\": stroke_index,\n    \"stroke_time\": total_time(stroke),\n    \"stroke_time_first\": first_time(stroke),\n    \"stroke_time_last\": last_time(stroke),\n    \"stroke_time_pause\": pause_before(drawing, stroke, stroke_index),\n    \"stroke_time_min\": min_time(stroke),\n    \"stroke_time_max\": max_time(stroke),\n    \"stroke_time_mixedup\": mixed_up_time(stroke)\n}\n\nreturn stats\n</code></pre>",
      "votes": 1,
      "replies": [
        {
          "id": 420206,
          "author_name": "Yiheng Wang",
          "author_url": "",
          "post_date": "2018-11-13T09:07:08.923000",
          "content": "<p>Hi Heng. Does the dataset you mentioned here in github the same as kaggle provide? Thus we have to download it for the stroke time information right?</p>",
          "votes": 0,
          "replies": []
        }
      ]
    },
    {
      "id": 425796,
      "author_name": "Leigh",
      "author_url": "",
      "post_date": "2018-11-22T05:51:35.180000",
      "content": "<p>I can't found this info for test data. Only country code</p>",
      "votes": 0,
      "replies": [
        {
          "id": 427051,
          "author_name": "Artur Ispiriants",
          "author_url": "",
          "post_date": "2018-11-24T12:44:08.350000",
          "content": "<p>It exists only in raw data, third item in each stroke array.</p>",
          "votes": 1,
          "replies": []
        }
      ]
    },
    {
      "id": 420250,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-13T10:51:16.803000",
      "content": "<p>encode using draw and pause time.</p>\n\n<p>note that the \"person\" become \"evident\". </p>\n\n<p>hence the correct label is likely to be a bed</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/420250/10668/0.40_9591733508469962_cooler_oven_diving_board.png\" alt=\"enter image description here\"></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 419281,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-11T16:21:30.477000",
      "content": "<p>using strokes for part attention:</p>\n\n<p>\"Weakly Supervised Local Attention Network for Fine-Grained Visual Classification\"\n- Tao Hu, Arvix 2018</p>\n\n<p>can also augment auxiliary features (e.g. stroke length, time, etc) to each part feature</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/419281/10656/attention_pooling.png\" alt=\"enter image description here\"></p>\n\n<hr>\n\n<p>related:</p>\n\n<p>interacting part features:</p>\n\n<p>\"Low-rank Bilinear Pooling for Fine-Grained Classification\"- Shu Kong, Arvix 2016</p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 418310,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-09T16:19:03.683000",
      "content": "<p>attention on the bridge:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10652/bridge1.png\" alt=\"enter image description here\">\n   <img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10653/bridge2.png\" alt=\"enter image description here\"></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 418302,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-09T16:07:08.883000",
      "content": "<p>here are more attention map:</p>\n\n<p><img src=\"https://storage.googleapis.com/kaggle-forum-message-attachments/418302/10651/encode_attent2.png\" alt=\"enter image description here\"></p>",
      "votes": 0,
      "replies": []
    },
    {
      "id": 418284,
      "author_name": "hengck23",
      "author_url": "",
      "post_date": "2018-11-09T15:36:01.527000",
      "content": "<p>check the raw data!</p>\n\n<p>the information is different. seems that raw data has timestamp per point?</p>",
      "votes": 0,
      "replies": [
        {
          "id": 418321,
          "author_name": "Guanshuo Xu",
          "author_url": "",
          "post_date": "2018-11-09T16:37:37.060000",
          "content": "<p>I think so. There is no one-on-one mapping between simplified and raw points, but there is one-on-one mapping for strokes. So if we choose to work on simplified CSVs, we can only use the time information stroke-wise. </p>",
          "votes": 0,
          "replies": []
        },
        {
          "id": 418961,
          "author_name": "Artyom Palvelev",
          "author_url": "",
          "post_date": "2018-11-11T01:04:40.097000",
          "content": "<p>Yes, the dataset description page on GitHab says so. </p>",
          "votes": 0,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "416284": "I read that quick draw dataset has \"draw time\" and\" pause time\" attribute. Is it true? \n\n![enter image description here][1]\n\n\n  [1]: https://i1.wp.com/teaching.statistics-is-awesome.org/wp-content/uploads/data-card.png",
    "418298": "using time to encode attention:\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/418298/10650/encode_time.png",
    "416384": "ok, i found it:\n\nhttps://github.com/vlandham/quickdraw/blob/master/analysis/all_aggregates.csv\n\nhttps://github.com/vlandham/quickdraw/blob/master/analysis/extract_times.ipynb\n\n\n\n\n    def first_time(stroke):\n         return stroke[2][0]\n\n...\n\n   def gen_stats_stroke(drawing, stroke, stroke_index):\n    \n    last_stroke = drawing['drawing'][len(drawing['drawing']) - 1]\n    stats = {\n        #\"key_id\": drawing['key_id'],\n        #\"recognized\": drawing['recognized'],\n        #\"word\": drawing['word'],\n        #\"countrycode\": drawing['countrycode'],\n        #\"drawing_time\": last_time(last_stroke),\n        #\"stroke_count\": len(drawing['drawing']),\n        \"stroke_index\": stroke_index,\n        \"stroke_time\": total_time(stroke),\n        \"stroke_time_first\": first_time(stroke),\n        \"stroke_time_last\": last_time(stroke),\n        \"stroke_time_pause\": pause_before(drawing, stroke, stroke_index),\n        \"stroke_time_min\": min_time(stroke),\n        \"stroke_time_max\": max_time(stroke),\n        \"stroke_time_mixedup\": mixed_up_time(stroke)\n    }\n    \n    return stats",
    "425796": "I can't found this info for test data. Only country code",
    "420250": "encode using draw and pause time.\n\nnote that the \"person\" become \"evident\". \n\nhence the correct label is likely to be a bed\n\n\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/420250/10668/0.40_9591733508469962_cooler_oven_diving_board.png",
    "419281": "using strokes for part attention:\n\n\"Weakly Supervised Local Attention Network for Fine-Grained Visual Classification\"\n- Tao Hu, Arvix 2018\n\ncan also augment auxiliary features (e.g. stroke length, time, etc) to each part feature\n\n   ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/419281/10656/attention_pooling.png\n\n\n---\n\nrelated:\n\ninteracting part features:\n\n\"Low-rank Bilinear Pooling for Fine-Grained Classification\"- Shu Kong, Arvix 2016",
    "418310": "attention on the bridge:\n\n\n   ![enter image description here][1]\n   ![enter image description here][2]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10652/bridge1.png\n  [2]: https://storage.googleapis.com/kaggle-forum-message-attachments/418310/10653/bridge2.png",
    "418302": "here are more attention map:\n\n\n  ![enter image description here][1]\n\n\n  [1]: https://storage.googleapis.com/kaggle-forum-message-attachments/418302/10651/encode_attent2.png",
    "418284": "check the raw data!\n\nthe information is different. seems that raw data has timestamp per point?"
  }
}