Intelligent Deep Learning Empowered Text Detection Model from Natural Scene Images

The scene Text Recognition process has become a hot research topic and challenging task owing to the complicated background, varying light intensities, colors, font styles, sizes. extraction from natural images encompasses two main processes: text detection recognition. latest advancements in Machine Learning (ML) Deep (DL) concepts can effectually automate recognition by training model properly. In this view, paper presents an Automated DL empowered Detection Natural Scene Images (ADLTD-NSI). ADLTD-NSI technique includes important Firstly, single shot detector (SSD) with Inception-v2 as baseline is employed for detection, object based on VGG-16 framework feature map followed six convolution layers. Secondly, Convolutional Recurrent Neural Network (CRNN) utilized process. Besides, recurrent layers CRNN utilize long short-term memory (LSTM) encoding sequence of vectors. Lastly, Connectionist Temporal Classification (CTC) loss applied predict labels equivalent sequences A wide range experiments was carried out benchmark COCO datasets, results are examined several aspects. experimental outcomes showcased better performance over other compared methods maximum accuracy 96.78%.