Project

Captcha Solver

A TensorFlow-based captcha solver trained on a custom labeled dataset for high-accuracy detection.

• Designed a captcha solver with TensorFlow models and achieved over 90 percent accuracy.
• Manually labeled 1,511 images with LabelImg to create the training dataset.
• Tested multiple TensorFlow Model Zoo architectures before selecting Faster RCNN Inception V2.
• Trained and deployed the model for real-world captcha solving workflows.
• Balanced model accuracy and inference efficiency for practical use.