{
  "fetched_at": "2026-10-19T07:19:24Z",
  "source_url": "https://www.kaggle.com/competitions/datathonindoml-2022/data",
  "pages": [
    {
      "name": "data-description",
      "content": "The dataset consists of 16000 grayscale images of documents such as letter, memo, email etc. The task is to predict the label given a document image. The performance will be evaluated based on mean F score. \n\n## Files\n\n*   **train** - this folder consists of the training images as <id>.tiff file. DO NOT change the ids of the individual image files as they are used for mapping to the corresponding classes.\n*   **validation** - this folder consists of 900 images as <id>.tiff file for the participants to check their performance. \n*   **train_labels.csv** - map of the images to the corresponding classes.  \n*   **metaData.csv** - supplemental information about the data\n\n## Columns\n\n*   `id` - image id || <id>part of the image file.|| E.g., if the image file 1234.tiff -> 1234\n*   `label` - class label for the corresponding image|| In total 16 classes indexed from 0 to 15",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "Given a set of grayscale document images, the task is to classify each image into one of the 16 classes or document types. The training dataset consists of 16000 images with 1000 images belonging to each class. The dataset was collected from the RVL-CDIP dataset [1].\n\n![](https://lh4.googleusercontent.com/g2CLiBvekM5cRzBCZA4zHDMOypCTmIYxfQIbjtDADEssJFiQ7FiOd5X5wsip2D4eCdJP54R2FyFI1sa-9YZK1ZM0ArP63Q--4J06x4y9038tZdiBmFAmTfPGEb_L-1rIZA=w1280)\n\n![](https://lh6.googleusercontent.com/_taiQGOKhNQwU_bkgDKMYTefotZe2ihA0Qec5Uy-9B9vi3kqLcHioQwetSdbcE3tK8iFCxSyfFRF5GMYKxK8RpztPnfDQpy2D6hOt9OtslbxPNlc3JRTo29woOR9QoPcuw=w1280)\n\n**Check out our Datathon [page](https://sites.google.com/view/datathonindoml/home) for more information**\n\n**Note: All the teams are requested to fill this [form](https://forms.gle/Dh2GQeuDAVvwFaNTA) to be eligible for prizes** (ignore, if already filled).\n\n**Note: All teams are requested to join this [group] (https://groups.google.com/g/datathonatindoml-2022). We will be making all the necessary announcements here.**\n\n**Note: We would like to remind all the participants that attaining the highest score on the public leaderboard does not guarantee a winning position. The submitted code will be tested on a separate private dataset. The merit of the work will also be judged based on the submitted report and other judging criteria as mentioned on the [Datathon webpage](https://sites.google.com/view/datathonindoml/home) . So, teams will have a chance even if the model does not attain the highest score on the specific validation set.**\n\n**For questions, reach out to us at <datathonatindoml@gmail.com>**\n\n[1] A. W. Harley, A. Ufkes, K. G. Derpanis, \"Evaluation of Deep Convolutional Nets for Document Image Classification and Retrieval,\" in ICDAR, 2015 \n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "The evaluation pages describes how submissions will be scored and how students should format their submissions. You don't need a sub-title at the top; the page title appears above. For an example of a high quality evaluation page, see the 2013 KDD Cup page, which we have included below:\n\n---\n\nThe evaluation metric for this competition is [Mean F1-Score](https://en.wikipedia.org/wiki/F-score). The F1 score, commonly used in information retrieval, measures accuracy using the statistics precision \\( \\text{p} \\) and recall \\( \\text{r} \\).\n\nPrecision is the ratio of true positives \\(\\text{tp}\\) to all predicted positives \\(\\text{tp} + \\text{fp}\\). Recall is the ratio of true positives \\(\\text{tp}\\) to all actual positives \\(\\text{tp} + \\text{fn}\\). The F1 score is given by:\n\n\\[ \\text{F1} = 2\\frac{\\text{p} \\cdot \\text{r}}{\\text{p}+\\text{r}}\\ \\ \\mathrm{where}\\ \\ \\text{p} = \\frac{\\text{tp}}{\\text{tp}+\\text{fp}},\\ \\ \\text{r} = \\frac{\\text{tp}}{\\text{tp}+\\text{fn}} \\]\n\nThe F1 metric weights recall and precision equally, and a good retrieval algorithm will maximize both precision and recall simultaneously. Thus, moderately good performance on both will be favored over extremely good performance on one and poor performance on the other.\n\n## Submission Format\n\nCheck out the sample submission file in Data tab",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "\n\n- Participants should work in a team of maximum 3 members. A Google Form will be circulated for registration of the teams. Once the team registers at Kaggle, there would be no further changes.\n\n- Each team should have at least one person from an Indian University or an Indian research lab.\n\n- Submission of code/implementation details and report is mandatory to be considered for prizes. In case a deep neural network based model is used, participants will be asked to upload the trained model.\n\n- The participants must only use the dataset provided for training and may not use the original larger dataset. The challenge is to propose a solution even with limited data. Using the original dataset might lead to disqualification. \n\n- The organizers will take the final call on the final prize money as well as any modification of the evaluation criteria (if any).\n\n- Don't cheat!\n\n- Apply yourself!\n\n- Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}