{
  "fetched_at": "2026-10-20T01:12:28Z",
  "source_url": "https://www.kaggle.com/competitions/eliva-26-image-restoration/data",
  "pages": [
    {
      "name": "data-description",
      "content": "**Dataset**\nIn the data26 folder, you can find a set of degraded images.\nInspect the images and devise a tool to estimate the degradation in order to remove it.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "The goal of this assignment is to restore degraded images using image processing techniques. \nThe degradation is unkonwn and must be estimated and corrected properly.\n\nThe challenge ranking will be done by estimating the restoration quality using the Peak Signal-to-Noise Ratio ([PSNR](https://en.wikipedia.org/wiki/Peak_signal-to-noise_ratio)) and Structural Similarity Index Measure ([SSIM](https://en.wikipedia.org/wiki/Structural_similarity_index_measure)) metrics.\n\nThe Data section contains the set of images to process. The goal is to improve their quality. You can start by inspecting the images to understand their nature and how they can be improved.\n\nYou can find a sample notebook with a dummy solution in the Code section. The script can be used to loop over all images in data26/, apply your enhancement solution, and create the .csv file to be submitted to the challenge. The enhanced images are encoded into the .csv that you upload as a solution.",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "The quality of each restored image in you subission will be ranked as follows:\n$$p_i= PSNR (\\textrm{restored}_i, \\textrm{original}_i) \\in [0, +\\infty]$$\n$$s_i =SSIM (\\textrm{restored}_i, \\textrm{original}_i) \\in [0,1]$$\nThe submission score is computed  as the average product of both metrics\n$$score = \\frac{1}{N} \\sum_i s_i p_i $$ \nA perfectly recovered image is expected to get \n$$s_i \\approx 1 $$ \n $$p_i > 20 \\ \\textrm{dB}$$ ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "rules",
      "content": "Don't cheat and have fun!",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "abstract",
      "content": "This is the first exercize in Eliva 2026 class.\nThe goals are:\n-  be familiar with kaggle env\n-  design an image restoration tool",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    }
  ]
}