{
  "fetched_at": "2026-10-22T10:29:18Z",
  "source_url": "https://www.kaggle.com/competitions/111-1-ntut-dl-app-hw4/data",
  "pages": [
    {
      "name": "data-description",
      "content": "This is NTUT 4K drone video dataset for human detection. We have use several commerical drones to record videos at different places.  \n\n## Files\n\n*   **ntut_drone_train.rar** - the training set\n*   **ntut_drone_test.rar** - the test set\n*   **labels** - The human bounding boxes are saved in csv format for each video\n\n## Columns\n\n*   `image` - video filename + frame ID\n*   `x_min` - x coordinate of upper left corner\n*   `y_min` -  y coordinate of upper left corner\n*   `x_max` - x coordinate of lower right corner\n*   `y_max` - y coordinate of lower right corner",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Description",
      "content": "In this homework we will try to detect humans in 4K drone videos. The human targets can be in different poses such as sitting, and under different view angles, e.g. bird's eye view (BEV).  ",
      "isPublished": false,
      "hasIsPublished": false,
      "postTitle": "",
      "hasPostTitle": false,
      "mimeType": "",
      "hasMimeType": false
    },
    {
      "name": "Evaluation",
      "content": "We only have one class, human, in our dataset. The quality of the bounding boxes will be evaluated by computing [Average Precision (AP)](https://en.wikipedia.org/wiki/Evaluation_measures_(information_retrieval)#Average_precision). However, it seems like the Kaggle Community Competition doesn't support calculating AP for object detection, we have to use Absolute Error (AE) instead. On this other hand, I think using F1 score is easier and see the influence of different confidence threshold. Therefore, you only need to submit your F1 score at your selected threshold. The Leaderboard will show the error (1 - your_AP). ",
      "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
    }
  ]
}