{
  "fetched_at": "2026-10-19T07:12:06Z",
  "source_url": "https://www.kaggle.com/competitions/evohackaton/data",
  "pages": [
    {
      "name": "rules",
      "content": "",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "data-description",
      "content": "<h2>\nData description\n</h2>\n\n<li>\n<strong>train.csv</strong> file has two columns – `name` and `category`.\n</li>\n\n`name` is the filename of an image. That image can be found in the `train` folder. \n\n`category` is the corresponding category of that image. There are 100 different categories.\n\n<li>\n<strong>train.zip</strong> file has images mentioned in <strong>train.csv</strong>\n</li>\n\n<li>\n<strong>test.zip</strong> file has images for which you need to predict a corresponding category.\n</li>\n\n<h2>\nGoal\n</h2>\n\nYour goal is to predict the correct classes for images.\n\nEach image belongs to 1 of 100 categories.\n\nFor each image in the test set you must predict an integer value in range `[0, 100)`.\n\n<h2>\nMetric\n</h2>\n\nSubmissions will be evaluated based on their <a href=\"https://en.wikipedia.org/wiki/Accuracy_and_precision#In_binary_classification\">accuracy</a> score.\n\n<h2>\nSubmission File Format\n</h2>\n\nYou should submit a `.csv` file with exactly 16858 rows plus a header row.\n\nHeader row should only have two columns called `name` and `category`.\n\nYour submission .csv file should have the following format:\n\n    name,category\n    0.jpg,9\n    1.jpg,19\n    2.jpg,29\n    ...etc",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "<h2>Welcome to the hackathon!</h2>\n\n<p><strong>This is a Transfer Learning Image Classification competition.</strong></p>\n<p>This challenge is dedicated to strengthen your skills in the Transfer Learning area.</p>\n<p>You are allowed (and it is highly recommended) to use any open-source pre-trained models in order to predict the classes of images.</p>\n\n<h2>Ideas and hints</h2>\n<p>Here are a few options which might be worth a try.</p>\n\n<p>\n<strong>1. Similar images might have similar image vectors</strong>\n</p>\n\n<li>\ndownload a pretrained `ImageNet` model\n</li>\n\n<li>\ndelete its bottom layer, which is responsible for final classification\n</li>\n\n<li>\nrun the `predict` function for images in `train.zip` and `test.zip` to get the image vectors\n</li>\n\n<li>\ncompare each image vector from `test.zip` versus all image vectors from `train.zip` (using cosine or euclidean distance)\n</li>\n\n<li>\nuse KNN to predict the category of an image\n</li>\n\n<p>\n<strong>2. Ensemble techniques might improve your results</strong>\n</p>\n\n<li>\ninstead of using one pretrained model, try using five different pretrained models\n</li>\n\n<li>\ninstead of predicting the class of an image directly, try averaging the predictions of rotated/transformed/shifted images\n</li>\n\n<p>\n<strong>3. Try data augmentation and image preprocessing techniques</strong>\n</p>\n\n<li>\ndownload a pretrained image segmentation model\n</li>\n\n<li>\nuse it to separate the content image from the background image\n</li>",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "<h2>\nGoal\n</h2>\n\nYour goal is to predict the correct classes for images.\n\nEach image belongs to 1 of 100 categories.\n\nFor each image in the test set you must predict an integer value in range `[0, 100)`.\n\n<h2>\nMetric\n</h2>\n\nSubmissions will be evaluated based on their <a href=\"https://en.wikipedia.org/wiki/Accuracy_and_precision#In_binary_classification\">accuracy</a> score.\n\n<h2>\nSubmission File Format\n</h2>\n\nYou should submit a `.csv` file with exactly 16858 rows plus a header row.\n\nHeader row should only have two columns called `name` and `category`.\n\nYour submission .csv file should have the following format:\n\n    name,category\n    0.jpg,9\n    1.jpg,19\n    2.jpg,29\n    ...etc",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}