A proof of concept Recaptcha solver using YOLOv3 on Tensorflow 2.0 and Selenium. This tutorial shows that with a better trained object detection weight file, ReCaptcha can be easily solved.
Install requirements and download pretrained weights:
$ pip3 install -r ./docs/requirements.txt
$ wget https://pjreddie.com/media/files/yolov3.weights
Download ChromeDriver and put it into $PATH.
In this part, we will use pretrained weights to make predictions on ReCaptcha.
$ python selenium_demo.py
- Click images using their pixel positions after finding objects and click VERIFY.
- Using NLP to find relation between "find buses" -> "bus" etc.
- Solve multiple screens such as "skip" or "until no objects left".