{
  "fetched_at": "2026-10-18T13:55:48Z",
  "source_url": "https://www.kaggle.com/competitions/applications-of-deep-learning-wustl-fall-2020/data",
  "pages": [
    {
      "name": "data-description",
      "content": "<p>The data consists of images of block structures.  Some of these structures are stable; while, others are unstable.  You must create a binary classifier that determines if a building is stable (1) or unstable (0).</p>\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/2.png\">\n\n<p>Not all images are it as well as others.  For example, the following image might be much more difficult due to lighting.</p>\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/108.png\">\n\n<p>Not all buildings will be of the same structure; consider the following image.</p>\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/314.png\">\n\n<p>The following blocks are used to create the buildings that you are predicting.  All blocks have different colors except for the two yellow blocks.  These are the only blocks that you will see in this data set. The colors define the shape of the blocks. \n All blocks of the same type will always be the same size.</p>\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/all-blocks.png\">\n\n<p>This overhead view shows the relative sizes of the blocks.</p>\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/block_sizes.png\">\n\n<h2>File descriptions</h2>\n<ul>\n<li><strong>train.csv</strong> - the training set</li>\n<li><strong>test.csv</strong> - the test set</li>\n<li><strong>sampleSubmission.csv</strong> - a sample submission file in the correct format</li>\n</ul>\n<h2>Data fields</h2>\n<ul>\n<li><strong>id</strong> - A unique ID for each image. This ID corresponds to the filename of each image.  For example, ID of 1 would be 1.png</li>\n<li><strong>stable</strong> - The label, 1 means a stable structure, 0 is unstable.</li>\n</ul>\n\n\n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "Determining if something is currently, and also likely to remain, structurally stable is a common engineering task.  Consider this water tower. Clearly it currently is not stable.  Rather, this tower is in free fall.  \n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/blocks-watertower.png\" width=\"320\">\n\nIn this Kaggle we will attempt to predict the structural stability of block buildings that I render with Blender and used simulated physics to determine the stability.  The data set is made up of many different buildings, built from a simulated set of wooden building blocks.  Your task is to predict if the building is stable or not.  Consider the following stable building.\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/blocks-stable.png\">\n\nIf I were to run the simulated physics in Blender, the appearance of this building would not change much.  There might be some minor settling; however, there would be no significant shift in the positions of the blocks.  However, not all of my block buildings are stable, consider the following unstable building.\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/blocks-unstable.png\">\n\nCan you see the source of the instability?  Look closely at the bottom row of blocks and how the four ceiling green blocks are supported.  If we run the physics simulation for a second, look at what happens.\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/blocks-falling.png\">\n\nClearly the building is not stable, it would nearly completly collapse.  Can you predict which buildings are stable just from then opening picture?\n\nThere are other complexities in the data as well.  Different images will have different angles and lighting of the buildings.\n\n<img src=\"https://data.heatonresearch.com/images/wustl/kaggle/2020-fall/blocks-samples.png\">\n\nAlso, not all block structures are the same basic shape.  However, all block structures are built from the same basic types of blocks.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "<p>The evaluation metric for this competition is <a href=\"http://wiki.fast.ai/index.php/Log_Loss\">Log Loss</a>. The log loss score, commonly used in binary classification, measures the performance of the predicted probability given by a model. The log loss score is given by:</p>\n<p>\\[ -{(y\\log(p) + (1 - y)\\log(1 - p))} \\]</p>\n<p>Logarithmic loss measures the performance of a classification model where the prediction input is a probability value between 0 and 1. The goal of our machine learning models is to minimize this value. A perfect model would have a log loss of 0. Log loss increases as the predicted probability diverges from the actual label. So predicting a probability of .012 when the actual observation label is 1 would result in a high log loss. </p>\n<h2>Submission Format</h2>\n\n<p>The file should contain a header and have the following format:</p>\n<pre>id,stable\n1,0.64\n2,0.64\n3,0.64\n4,0.64\n5,0.64\n6,0.64\n7,0.64\n8,0.64\n...</pre>\n\n<h2>Rules</h2>\n\n* Teams may be up to 5 members\n* You may not human-label the test data, your predictions must come from a model, not from a human manually labeling the images.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "This is a page where you can include rules that participants must accept before joining. You may wish to include rules like:\n\n - Don't cheat!\n - Apply yourself!\n - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}