{
  "fetched_at": "2026-10-19T10:48:39Z",
  "source_url": "https://www.kaggle.com/competitions/elte-traffic-signs/data",
  "pages": [
    {
      "name": "data-description",
      "content": "# Data description\n\nAll files required are attached in the data section of the challenge.\n\n## What am I predicting?  \n\nYou are predicting a row in the form: id, class,x1,y1,x2,y2 for each image where x1,y1,x2 and y2 are the corner coordinates in pixels of the bounding box of the object to localize. \n \n## Files provided\n\n*   **train.csv** - The training set: A set of image paths and ground truth labels for class and bounding box.\n*   **test.csv** - The test set: A set of image paths without ground truth predictions.\n*   **sample.csv** - A sample submission file in the correct format. This shows you how your submission file should look.\n\n## Columns\n*   `id` - Identification of png image in dataset.\n*   `new_path` - Path to png image in dataset.\n*   `class` - Number from 0-4 that defines the class of the sign.  \n*   `x1` -  X coordinate of bounding bottom box corner. \n*   `y1` -  Y coordinate of bounding bottom box corner. \n*   `x2` -  X coordinate of bounding top box corner. \n*   `y2` -  Y coordinate of bounding top box corner. ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "# Traffic sign localization\n\n![Meme](https://i.imgflip.com/68uh3g.jpg)\n\nWith the advance of self driving car technology, the demand for high quality computer vision systems is increasing. On roads, different signs offer important information to a human driver through their visual system.\n \nFor a self driving system we gather this information with a camera and process it with the help of deep learning and computer vision. This could be very useful as a self driving system can be made more aware of the context of the area around it. This has impacts on car safety as well as even navigation where no GPS signal is available. \n\n![Self driving](https://nypost.com/wp-content/uploads/sites/2/2021/08/elon-musk-tesla.jpg?quality=80&strip=all)\n\nIn order to process road sign information the computer system must both localize the sign in space i.e. get the `[x1,y1,x2,y2]` pixel coordinates that define where the sign is in the image\nand perform a classification in order to define what the meaning of the sign is. \nIn Machine Learning, the \"type\" of object is referred to as a 'class'.\n\nIn this challenge the dataset contains 676 images of 4 separate classes: \n` {'speedlimit': 0, 'stop': 1, 'crosswalk': 2, 'trafficlight': 3}`\nwith each number in the above dictionary representing its human understandable string.\n\n![Classification](https://www.brandonindustries.com/wp-content/uploads/2019/04/Road_Signs.jpg)\n\nThe challenge is based around the concept of performing both Regression and Classification simultaneously.  The regression is to predict the bounding box coordinates of the sign and the classification to find its class. \n\nThis challenge involves: image pre-processing ,building and training a deep neural network.\n\nThe testing will be done on Kaggle with 60 test images\n\nA secret held out test set containing 200 images will be used for the final leader board.\n\n# Tips. \n\nMake sure you load the data properly: plot the bounding boxes on the images for the training set to make sure they are correctly aligned.\n\n# Loading .csv files for use. \ndf_train = pd.read_csv('./train.csv',delimiter=',',converters={'new_bb': literal_eval})\n```\n\nRemember: some augmentations must be performed both to the image and its bounding box!\n\nTransfer learning - 676 images is not a lot so your models may benefit from some prior knowledge ...\n\nHave fun! \n\n## Acknowledgements\n\nD. Ghica, Si Wei Lu and Xiaobu Yuan, \"Recognition of traffic signs by artificial neural network,\" Proceedings of ICNN'95 - International Conference on Neural Networks, 1995, pp. 1444-1449 vol.3, doi: 10.1109/ICNN.1995.487372.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "# How is the competition evaluated?\n\nThe evaluation metric for this competition is the [Root mean squared error: RMSE ](https://en.wikipedia.org/wiki/Root-mean-square_deviation). The RMSE  will measure both classification and regression deviation from the ground truth. \n\n## Submission Format\n\nAttached is an example file sample.csv which outlines the exact format the challenge accepts.\nThe solutions file MUST be 200 rows long with a header row as shown in the sample file.\n\nHeader format: \n```,class,x1,y1,x2,y2```\n\nRows example format:\n```0,0.0,107.39075,172.88014,155.44627,240.57771```\n\nThe first id column must be in integer format and the id's must match the id's of the test set to be correctly evaluated. \n\nEvery other value can be in integer or float format. \n",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": " - Don't cheat!\n - Apply yourself!\n - Have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}