{
  "id": 42283,
  "title": "parallel execution on spark cluster",
  "url": "/competitions/cdiscount-image-classification-challenge/discussion/42283",
  "author_name": "matthew leung",
  "post_date": "2017-10-29T05:46:48.510000",
  "votes": 0,
  "comment_count": 0,
  "views": 0,
  "content": "<p>I used VGG16 to classify the image to the feature vector.  Running on the spark cluster with 49 servers in google cloud, it can achieve 47 images per second, while the through it just 0.6 if it run on a single node.  It can apply to other image recognition models.  If any would like to speed up his model, we can join team and share the cost of google cloud together.</p>",
  "messages": [
    {
      "id": 237016,
      "postDate": "2017-10-29T05:46:48.510Z",
      "content": "<p>I used VGG16 to classify the image to the feature vector.  Running on the spark cluster with 49 servers in google cloud, it can achieve 47 images per second, while the through it just 0.6 if it run on a single node.  It can apply to other image recognition models.  If any would like to speed up his model, we can join team and share the cost of google cloud together.</p>",
      "rawMarkdown": "I used VGG16 to classify the image to the feature vector.  Running on the spark cluster with 49 servers in google cloud, it can achieve 47 images per second, while the through it just 0.6 if it run on a single node.  It can apply to other image recognition models.  If any would like to speed up his model, we can join team and share the cost of google cloud together."
    }
  ],
  "comments": [],
  "raw_markdown_by_id": {
    "237016": "I used VGG16 to classify the image to the feature vector.  Running on the spark cluster with 49 servers in google cloud, it can achieve 47 images per second, while the through it just 0.6 if it run on a single node.  It can apply to other image recognition models.  If any would like to speed up his model, we can join team and share the cost of google cloud together."
  }
}