Asad-Ismail / CaptchaRecognition

Captcha recognition using deep learning and automated appointment time check

Captcha Recognition Dataset/Model and automation

This project presents a custom Captcha dataset and deep learning model for Captcha recognition, along with a script to automate appointment systems using the trained model.

Key Contributions

• A custom Captcha dataset with 6 characters per captcha
• A deep learning model for Captcha recognition
• Deployment of the model to read Captchas and automate appointment systems

Requirements

1. Install Chrome using the provided installchrome.sh script.
2. Install a compatible Chrome driver. A driver is included in the repository, but if it is incompatible, download and install the appropriate version.
3. Install the required packages using: pip install -r requirements.txt.

Dataset

You can download the 5 digit Captcha dataset from link below

Few examples of the dataset are shown below

       

       

       

After training on 5 digit dataset we further finetune the model on 6 digit Captcha dataset

Datasets and trained models for (6 char captcha) can be downloaded from below

https://drive.google.com/drive/folders/1NS9CBTT-bu7Uegc6t2jn3Xg3m8iUtsZz?usp=share_link

Model Training

Refer to the train_captcha.ipynb notebook for instructions on training the Captcha recognition model with the provided dataset. The model architecture utilizes two bidirectional LSTMs and CTC loss for Captcha recognition. It has low latency and demonstrates high accuracy in online tests.

Automatic Appointment System

This is the only thing you need if you just want to crack captcha to get appointment The automation_germany.py or automation_vfs.py script leverages the trained Captcha recognition model to identify Captchas on websites and verify the availability of appointments. If an appointment is available, it sends an email notification to the specified recipient. While the automation is tailored to a specific type of website, it can serve as a blueprint for creating appointment or alert systems for other websites and services.

Note!! Change links and email_sender password and email_receivers to the desired values in automation_germany.py