We need moar candy canes!?

build_capteha_images.md

Need more images from the Frido Sleigh CAPTEHA? Just want to test your training and categorization?

• Open the site in a Chrome browser and open the developer tools
• Log all the images through multiple runs of the CAPTEHA
• Capture images into a .har archive -- It's just json, and the request strings include the full image in base64 mode.
• Run the attached script to convert the images from the .har file to .png

Then run Chris Davis's scripts from https://github.com/chrisjd20/img_rec_tf_ml_demo to categorize your images!

build_capteha_images.py

import json
import base64

with open('fridosleigh.com.har') as f:
    har = json.load(f)

for imageset in range(0, len(har['log']['entries']):
    id = har['log']['entries'][imageset]['response']['content']['text']
    images = json.loads(id)['images']
    for image in range(0, len(images)):
        filename = f'set{imageset:02}image{image:03}.png'
        outfile = open(filename, 'wb')
        outfile.write(base64.b64decode(images[image]['base64']))
        outfile.close()