{
  "fetched_at": "2026-10-19T09:13:54Z",
  "source_url": "https://www.kaggle.com/competitions/comp551-modified-mnist/data",
  "pages": [
    {
      "name": "rules",
      "content": "##One account per participant\nYou cannot sign up to Kaggle from multiple accounts and therefore you cannot submit from multiple accounts.\n\n## No private sharing outside teams\n    \nPrivately sharing code or data outside of teams is not permitted.\n\nIt's okay to share code if made available to all participants on the forums.\n\n##Team Mergers\n\nTeam mergers are not allowed in this competition.\n\n##Team Limits\nThe  maximum team size is 3.\n\n## Submission Limits\n\nYou may submit a maximum of 2 entries per day.\nYou may select up to 2 final submissions for judging.\n\n##Competition Timeline\n  \nStart Date: **25/10/2017**\n\n\nEnd Date:  **13/11/2017 11:59 PM EDT**\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "data-description",
      "content": "The training set (train_x.csv, test_x.csv) consists of 50,000  8bit-greyscale images of size 64x64. There is a unique label for each of the examples, which is the result  of the application of operator on the two digits in the image ( result for \"a\"/\"A\" is the sum of two digits and result for \"m\"/\"M\"  is the product of two digits ). The total num of unique classes are 40 . ( [0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 20, 21, 24, 25, 27, 28, 30, 32, 35, 36, 40, 42, 45, 48, 49, 54, 56, 63, 64, 72, 81] )\n\n<p>\nThe test set (test_x.csv) consists of 10000 images of the same format.\n</p>\n\n\n<p>\nThe .csv file contains the raw uncompressed images pixel values in row major format separated by \",\". Each image is 64 x 64, so each row contains 4096 float values, each containing the a float value which represents the 0-255 greyscale intensity of each pixel. \n</p>\n\n\n\n\nHere's how to load (and view) this data in numpy format \n\n    import numpy   as np \n    import scipy.misc # to visualize only  \n    x = np.loadtxt(\"train_x.csv\", delimiter=\",\") # load from text \n    y = np.loadtxt(\"train_y.csv\", delimiter=\",\") \n    x = x.reshape(-1, 64, 64) # reshape \n    y = y.reshape(-1, 1) \n    scipy.misc.imshow(x[0]) # to visualize only \n\n\n**If you need help loading the data (in python or other languages), please post on MyCourses.** We will do our best to respond in a timely fashion.\n\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "<h2> Modified MNIST Challenge </h2>\n<p>For project #3, you are asked to create classifiers to analyze images consisting of two digits and a character. The character may be an 'a'  or 'A' (for add), in which case your classifier has to return the sum of the two digits,  or an 'm'  or 'M' (for multiply), in which case your classifier has to return the product of the two digits. Additionally, a variety of different textures are used in the backgrounds of the images.</p>\n<p>We provide a set of labeled training data (images and target outputs) as well as a set of unlabeled test data (i.e. images only).  You must run your model on the test data and submit the target outputs to Kaggle. </p>\n<p>You are allowed to use libraries such as theano/sklearn/tensorflow/pytorch <b>for parts 3 and 1 only</b>. For part 2 you must do the full implementation. You can submit any part for kaggle, although the kaggle competition is ranked based on your best submission only and you have a daily limit on number of submissions.</p>\n<p>Please keep the kaggle teams the same as the teams for the report, and include your kaggle team name in the report so we can grade your performance  :) </p>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "You will be evaluated based on classification accuracy on the private test set. The baseline performance of a random classifier is less than 2%. For submission formats, see the Data section.\n\n\n## Submission File Format\n\nThe file should contain a header and have the following format:\n\n```\nId,Label    \n1,0    \n2,0    \n3,0    \netc.   \n```\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}