In this competition, the participants have to decode the images containing jumbled text, such images are called CAPTCHAs. It is an opportunity for development of machine learning models that aim at recognizing and reading these distorted letters.


### The Dataset

Each image, of 160px width and 60px height, contains 5 alphanumerical characters - that are squeezed, distorted, noisy, etc. etc. These images were generated through the [`captcha` module in Python](https://github.com/lepture/captcha), as follows:

```py
from captcha.image import ImageCaptcha

import random
import string

def generate_captcha(text_length=5, name=0, kind='train'):
    image = ImageCaptcha(width=160, height=60)
    captcha_text = ''.join(random.choices(string.ascii_uppercase + string.digits, k=text_length))
    image_data = image.generate_image(captcha_text)

    image_data.save(f'../data/{kind}/{name}.png')
    return captcha_text, image_data
```