{
  "id": 101224,
  "title": "Kuzushiji Recognition Starter",
  "url": "/competitions/kuzushiji-recognition/discussion/101224",
  "author_name": "",
  "post_date": "2019-07-24T05:51:02.474649800Z",
  "votes": 5,
  "comment_count": 3,
  "views": 0,
  "content": "<h2><a href=\"https://www.kaggle.com/jesucristo/kuzushiji-recognition-starter\">Kuzushiji Recognition Starter</a></h2>\n\n<p>In this kernel you'll find:</p>\n\n<p><strong>1. EDA</strong>\nNew df_train\nmissing data\nchar stats\ntop-10 chars\ntop-100 chars (plot)\n<strong>2. Simple Visualization</strong>\n*<em>3. KMINST</em>*\nSave the 683464 chars/digits images in kminst.zip and info.csv\nExamples of obtained chars from a random image.\n<strong>4. KMINST Classification</strong>\nSimple KNN\nDeep Learning\n<strong>5. Simple Predictions Visualization</strong></p>\n\n<hr>\n\n<p><em>kernels update frequency = weekly</em></p>\n\n<p><img src=\"https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/__results___files/__results___23_0.png\" alt=\"\"></p>\n\n<p>![](<a href=\"https://i.ibb.co/MVfhLSP/Captura-de-pantalla-de-2019-07-24-07-40-46.png\">https://i.ibb.co/MVfhLSP/Captura-de-pantalla-de-2019-07-24-07-40-46.png</a> =400x*)</p>\n\n<p>![](results_files/<strong>results</strong>_44_58.png\"&gt;https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/<strong>results</strong>_files/<strong>results</strong>_44_58.png =400x*)</p>\n\n<p>&gt; You can download the 683464 chars/digits , check <strong>kminst.zip</strong> (kernels <code>output</code>)</p>\n\n<h3>My approach</h3>\n\n<p>Phase 1: character segmentation (ocr?)\nPhase 2: character classification. (check <a href=\"https://www.kaggle.com/c/kuzushiji-recognition/discussion/100579#latest-580915\">Must-read material</a>)</p>\n\n<h3>Next Step</h3>\n\n<p>&gt; I'm focused in other comps, you'll se this in the following days ...</p>\n\n<ul>\n<li>Digitalize 683464 digits (32 * 32 size of the gray-scale images). Create Train and Validation.</li>\n<li>KNN</li>\n<li>SVC (linear, rbf)</li>\n<li>CNN</li>\n<li>MobileNet ¿?</li>\n</ul>",
  "messages": [
    {
      "id": "583163",
      "postDate": "07/24/2019 05:51:02",
      "content": "<h2><a href=\"https://www.kaggle.com/jesucristo/kuzushiji-recognition-starter\">Kuzushiji Recognition Starter</a></h2>\n\n<p>In this kernel you'll find:</p>\n\n<p><strong>1. EDA</strong>\nNew df_train\nmissing data\nchar stats\ntop-10 chars\ntop-100 chars (plot)\n<strong>2. Simple Visualization</strong>\n*<em>3. KMINST</em>*\nSave the 683464 chars/digits images in kminst.zip and info.csv\nExamples of obtained chars from a random image.\n<strong>4. KMINST Classification</strong>\nSimple KNN\nDeep Learning\n<strong>5. Simple Predictions Visualization</strong></p>\n\n<hr>\n\n<p><em>kernels update frequency = weekly</em></p>\n\n<p><img src=\"https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/__results___files/__results___23_0.png\" alt=\"\"></p>\n\n<p>![](<a href=\"https://i.ibb.co/MVfhLSP/Captura-de-pantalla-de-2019-07-24-07-40-46.png\">https://i.ibb.co/MVfhLSP/Captura-de-pantalla-de-2019-07-24-07-40-46.png</a> =400x*)</p>\n\n<p>![](results_files/<strong>results</strong>_44_58.png\"&gt;https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/<strong>results</strong>_files/<strong>results</strong>_44_58.png =400x*)</p>\n\n<p>&gt; You can download the 683464 chars/digits , check <strong>kminst.zip</strong> (kernels <code>output</code>)</p>\n\n<h3>My approach</h3>\n\n<p>Phase 1: character segmentation (ocr?)\nPhase 2: character classification. (check <a href=\"https://www.kaggle.com/c/kuzushiji-recognition/discussion/100579#latest-580915\">Must-read material</a>)</p>\n\n<h3>Next Step</h3>\n\n<p>&gt; I'm focused in other comps, you'll se this in the following days ...</p>\n\n<ul>\n<li>Digitalize 683464 digits (32 * 32 size of the gray-scale images). Create Train and Validation.</li>\n<li>KNN</li>\n<li>SVC (linear, rbf)</li>\n<li>CNN</li>\n<li>MobileNet ¿?</li>\n</ul>",
      "rawMarkdown": "## [Kuzushiji Recognition Starter](https://www.kaggle.com/jesucristo/kuzushiji-recognition-starter)\n\nIn this kernel you'll find:\n\n**1. EDA**\nNew df_train\nmissing data\nchar stats\ntop-10 chars\ntop-100 chars (plot)\n**2. Simple Visualization**\n**3. KMINST**\nSave the 683464 chars/digits images in kminst.zip and info.csv\nExamples of obtained chars from a random image.\n**4. KMINST Classification**\nSimple KNN\nDeep Learning\n**5. Simple Predictions Visualization**\n\n----\n\n*kernels update frequency = weekly*\n\n![](https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/__results___files/__results___23_0.png)\n\n![](https://i.ibb.co/MVfhLSP/Captura-de-pantalla-de-2019-07-24-07-40-46.png =400x*)\n\n![](https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/__results___files/__results___44_58.png =400x*)\n\n&gt; You can download the 683464 chars/digits , check **kminst.zip** (kernels ```output```)\n\n### My approach\n\nPhase 1: character segmentation (ocr?)\nPhase 2: character classification. (check [Must-read material](https://www.kaggle.com/c/kuzushiji-recognition/discussion/100579#latest-580915))\n\n### Next Step\n\n&gt; I'm focused in other comps, you'll se this in the following days ...\n\n- Digitalize 683464 digits (32 * 32 size of the gray-scale images). Create Train and Validation.\n- KNN\n- SVC (linear, rbf)\n- CNN\n- MobileNet ¿?",
      "votes": null
    },
    {
      "id": "583844",
      "postDate": "07/25/2019 04:42:58",
      "content": "<p>Google vision api for ocr hmmm..?</p>",
      "rawMarkdown": "Google vision api for ocr hmmm..?",
      "votes": null
    },
    {
      "id": "583911",
      "postDate": "07/25/2019 06:50:46",
      "content": "<p>Could be nice, I'm working on something similar Genos ;) </p>",
      "rawMarkdown": "Could be nice, I'm working on something similar Genos ;)",
      "votes": null
    },
    {
      "id": "584044",
      "postDate": "07/25/2019 11:09:45",
      "content": "<p>I will be wait for that👍 :)</p>",
      "rawMarkdown": "I will be wait for that👍 :)",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 583844,
      "author_name": "vaishvik25",
      "author_url": "",
      "post_date": "07/25/2019 04:42:58",
      "content": "<p>Google vision api for ocr hmmm..?</p>",
      "votes": null,
      "replies": [
        {
          "id": 583911,
          "author_name": "jesucristo",
          "author_url": "",
          "post_date": "07/25/2019 06:50:46",
          "content": "<p>Could be nice, I'm working on something similar Genos ;) </p>",
          "votes": null,
          "replies": []
        },
        {
          "id": 584044,
          "author_name": "vaishvik25",
          "author_url": "",
          "post_date": "07/25/2019 11:09:45",
          "content": "<p>I will be wait for that👍 :)</p>",
          "votes": null,
          "replies": []
        }
      ]
    }
  ],
  "raw_markdown_by_id": {
    "583163": "## [Kuzushiji Recognition Starter](https://www.kaggle.com/jesucristo/kuzushiji-recognition-starter)\n\nIn this kernel you'll find:\n\n**1. EDA**\nNew df_train\nmissing data\nchar stats\ntop-10 chars\ntop-100 chars (plot)\n**2. Simple Visualization**\n**3. KMINST**\nSave the 683464 chars/digits images in kminst.zip and info.csv\nExamples of obtained chars from a random image.\n**4. KMINST Classification**\nSimple KNN\nDeep Learning\n**5. Simple Predictions Visualization**\n\n----\n\n*kernels update frequency = weekly*\n\n![](https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/__results___files/__results___23_0.png)\n\n![](https://i.ibb.co/MVfhLSP/Captura-de-pantalla-de-2019-07-24-07-40-46.png =400x*)\n\n![](https://www.kaggleusercontent.com/kf/17666491/eyJhbGciOiJkaXIiLCJlbmMiOiJBMTI4Q0JDLUhTMjU2In0..ujWt65G1GS7XInJuZ4eVVw.9WyKeP2k3mBmgpwu2xtwN3kuW89Sm22aBBLfhdbG1Ds3EZtCd51YnMLFV_AzWWVYuQ_WdNR3zk0ohxpPLc382VeBanDTFiyOHJNHvZMMJtChhoaCxE9c9ro162j6jb5RGnGOWpGEfC5E5Va43WdgVXV03h8RdhkOlLKo_wLx8B8.4Y6tF6CN4TxoOzBSGd_cXg/__results___files/__results___44_58.png =400x*)\n\n&gt; You can download the 683464 chars/digits , check **kminst.zip** (kernels ```output```)\n\n### My approach\n\nPhase 1: character segmentation (ocr?)\nPhase 2: character classification. (check [Must-read material](https://www.kaggle.com/c/kuzushiji-recognition/discussion/100579#latest-580915))\n\n### Next Step\n\n&gt; I'm focused in other comps, you'll se this in the following days ...\n\n- Digitalize 683464 digits (32 * 32 size of the gray-scale images). Create Train and Validation.\n- KNN\n- SVC (linear, rbf)\n- CNN\n- MobileNet ¿?",
    "583844": "Google vision api for ocr hmmm..?",
    "583911": "Could be nice, I'm working on something similar Genos ;)",
    "584044": "I will be wait for that👍 :)"
  },
  "source": "meta"
}