{
  "id": 396532,
  "title": "Problem Approach Sharing",
  "url": "/competitions/benetech-making-graphs-accessible/discussion/396532",
  "author_name": "",
  "post_date": "2023-03-22T00:54:43.504454300Z",
  "votes": 4,
  "comment_count": 2,
  "views": 0,
  "content": "<p>Am I correct in understanding the way to approach this competition is an image detection problem using graph images as training data and graph annotations as labels? </p>\n<p>Or could some of the annotation data be used along with images for training as we are only predicting chart x and y ticks and chart type?</p>\n<p>Share your thoughts on how to approach this competition!</p>",
  "messages": [
    {
      "id": "2191424",
      "postDate": "03/22/2023 00:54:43",
      "content": "<p>Am I correct in understanding the way to approach this competition is an image detection problem using graph images as training data and graph annotations as labels? </p>\n<p>Or could some of the annotation data be used along with images for training as we are only predicting chart x and y ticks and chart type?</p>\n<p>Share your thoughts on how to approach this competition!</p>",
      "rawMarkdown": "Am I correct in understanding the way to approach this competition is an image detection problem using graph images as training data and graph annotations as labels? \n\nOr could some of the annotation data be used along with images for training as we are only predicting chart x and y ticks and chart type?\n\nShare your thoughts on how to approach this competition!",
      "votes": null
    },
    {
      "id": "2191497",
      "postDate": "03/22/2023 02:56:38",
      "content": "<p>I think it's more of the latter. We use the annotation to help training our models to identify components in an image. Once we have finished training, we use it to identify images in test dataset.</p>",
      "rawMarkdown": "I think it's more of the latter. We use the annotation to help training our models to identify components in an image. Once we have finished training, we use it to identify images in test dataset.",
      "votes": null
    },
    {
      "id": "2192435",
      "postDate": "03/22/2023 16:46:23",
      "content": "<p>For text labels (from axis) use OCR models to extract it. As for the next steps, I'm still thinking.</p>",
      "rawMarkdown": "For text labels (from axis) use OCR models to extract it. As for the next steps, I'm still thinking.",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2191497,
      "author_name": "iqbalsyahakbar",
      "author_url": "",
      "post_date": "03/22/2023 02:56:38",
      "content": "<p>I think it's more of the latter. We use the annotation to help training our models to identify components in an image. Once we have finished training, we use it to identify images in test dataset.</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2192435,
      "author_name": "manwithaflower",
      "author_url": "",
      "post_date": "03/22/2023 16:46:23",
      "content": "<p>For text labels (from axis) use OCR models to extract it. As for the next steps, I'm still thinking.</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2191424": "Am I correct in understanding the way to approach this competition is an image detection problem using graph images as training data and graph annotations as labels? \n\nOr could some of the annotation data be used along with images for training as we are only predicting chart x and y ticks and chart type?\n\nShare your thoughts on how to approach this competition!",
    "2191497": "I think it's more of the latter. We use the annotation to help training our models to identify components in an image. Once we have finished training, we use it to identify images in test dataset.",
    "2192435": "For text labels (from axis) use OCR models to extract it. As for the next steps, I'm still thinking."
  },
  "source": "meta"
}