This project uses Deeplearning techiniques to break Google reCAPTCHA v2 by classifying the most popular image categories
We use a pretrained Efficient Net B0 model to classify 11 categories of Google reCAPTCHA v2-
'Bicycle', 'Bridge', 'Bus', 'Car', 'Chimney', 'Crosswalk', 'Hydrant', 'Motorcycle', 'Mountain', 'Palm', 'Traffic Light'
We finetune the pretrained model using manually scrapped images (about 1000 per class) using transfer learning
- Create a Virtual Environment
conda create -n captcha- Activate the virtual environment
conda activate captcha- Using command line navigate to the project directory
- Install the required packages
pip install -r requirements.txt- Add your custom image to Images folder
- Run the following command
streamlit run ImageCaptcha.py- Program will run in your local browser
