{
  "id": 487758,
  "title": "What is the input of the model",
  "url": "/competitions/geolifeclef-2024/discussion/487758",
  "author_name": "",
  "post_date": "2024-03-30T13:59:23.026174300Z",
  "votes": null,
  "comment_count": 2,
  "views": 0,
  "content": "<p>I am new to such a big competition. However, I want to train different models for different types of data sets. However, how will the inputs of the model be to predict like in Submission CVS?</p>",
  "messages": [
    {
      "id": "2723796",
      "postDate": "03/30/2024 13:59:23",
      "content": "<p>I am new to such a big competition. However, I want to train different models for different types of data sets. However, how will the inputs of the model be to predict like in Submission CVS?</p>",
      "rawMarkdown": "I am new to such a big competition. However, I want to train different models for different types of data sets. However, how will the inputs of the model be to predict like in Submission CVS?",
      "votes": null
    },
    {
      "id": "2725187",
      "postDate": "03/31/2024 11:53:29",
      "content": "<p>In my model, I extract the surveyID in PA and PO to get the image Patches and Timeseries, and use the coordinates to get the image of other Environment data. I use them as input to the model and a Linear layer is used at the end of the model to obtain prediction results. For the results of different models, I think this involves some multi-modal fusion. But the simplest method is directly connect the features extracted by different models together (but the effect may not be very good), then just use a Linear layer to get the predict. I think the work of <a href=\"https://www.kaggle.com/code/gobyeonggeon/preprocess-visualize-spatial-data-eda-xgb\" target=\"_blank\">preprocess-visualize-spatial-data-eda-xgb</a> is very helpful. Thanks to his/her contribution!</p>",
      "rawMarkdown": "In my model, I extract the surveyID in PA and PO to get the image Patches and Timeseries, and use the coordinates to get the image of other Environment data. I use them as input to the model and a Linear layer is used at the end of the model to obtain prediction results. For the results of different models, I think this involves some multi-modal fusion. But the simplest method is directly connect the features extracted by different models together (but the effect may not be very good), then just use a Linear layer to get the predict. I think the work of [preprocess-visualize-spatial-data-eda-xgb](https://www.kaggle.com/code/gobyeonggeon/preprocess-visualize-spatial-data-eda-xgb) is very helpful. Thanks to his/her contribution!",
      "votes": null
    },
    {
      "id": "2725559",
      "postDate": "03/31/2024 16:17:39",
      "content": "<p>Thank you very much. All my questions were answered</p>",
      "rawMarkdown": "Thank you very much. All my questions were answered",
      "votes": null
    }
  ],
  "comments": [
    {
      "id": 2725187,
      "author_name": "xuanchengx",
      "author_url": "",
      "post_date": "03/31/2024 11:53:29",
      "content": "<p>In my model, I extract the surveyID in PA and PO to get the image Patches and Timeseries, and use the coordinates to get the image of other Environment data. I use them as input to the model and a Linear layer is used at the end of the model to obtain prediction results. For the results of different models, I think this involves some multi-modal fusion. But the simplest method is directly connect the features extracted by different models together (but the effect may not be very good), then just use a Linear layer to get the predict. I think the work of <a href=\"https://www.kaggle.com/code/gobyeonggeon/preprocess-visualize-spatial-data-eda-xgb\" target=\"_blank\">preprocess-visualize-spatial-data-eda-xgb</a> is very helpful. Thanks to his/her contribution!</p>",
      "votes": null,
      "replies": []
    },
    {
      "id": 2725559,
      "author_name": "ahmetkoraysonal",
      "author_url": "",
      "post_date": "03/31/2024 16:17:39",
      "content": "<p>Thank you very much. All my questions were answered</p>",
      "votes": null,
      "replies": []
    }
  ],
  "raw_markdown_by_id": {
    "2723796": "I am new to such a big competition. However, I want to train different models for different types of data sets. However, how will the inputs of the model be to predict like in Submission CVS?",
    "2725187": "In my model, I extract the surveyID in PA and PO to get the image Patches and Timeseries, and use the coordinates to get the image of other Environment data. I use them as input to the model and a Linear layer is used at the end of the model to obtain prediction results. For the results of different models, I think this involves some multi-modal fusion. But the simplest method is directly connect the features extracted by different models together (but the effect may not be very good), then just use a Linear layer to get the predict. I think the work of [preprocess-visualize-spatial-data-eda-xgb](https://www.kaggle.com/code/gobyeonggeon/preprocess-visualize-spatial-data-eda-xgb) is very helpful. Thanks to his/her contribution!",
    "2725559": "Thank you very much. All my questions were answered"
  },
  "source": "meta"
}