# DL4CV Aircraft Detection — Data

来源：https://www.kaggle.com/competitions/dl-4-cv-aircraft/data

本地 Meta Kaggle 快照；不是本次实时页面。

# Dataset Description

Our dataset comes in the **YOLO** format

*From label-studio*: Popular TXT format is created for each image file. Each txt file contains annotations for the corresponding image file, that is object class, object coordinates, height & width.

For example *images/airplane_17.jpg* could come with a text file *labels/airplane_17.txt* which looks like
```csv
0 0.648333 0.470238 0.563333 0.869048
0 0.650000 0.023810 0.000000 0.023810
```
where rows denote bounding boxes and columns are `class` followed by normalized `center-x, center-y, width, height`.
You can find class label mappings in `notes.json`.