This project implements scene text detection and recognition using deep neural networks. It combines three models: a VGG16-based classifier for text orientation detection (0, 90, 180, 270 degrees), a CTPN (CNN+RNN) network for text region detection, and a CRNN (CNN+GRU/LSTM+CTC) network for end-to-end variable-length text recognition. The system supports both GPU and CPU environments with Python 3.6+ and TensorFlow 1.7.