A novel pipeline framework for multi oriented scene text image detection and recognition

• Published in: Expert systems with applications Vol. 170; p. 114549
• Main Authors: Naiemi, Fatemeh, Ghods, Vahid, Khalesi, Hassan
• Format: Journal Article
• Language: English
• Published: New York Elsevier Ltd 15.05.2021; Elsevier BV
• Subjects: Algorithms; Artificial neural networks; Autonomous cars; Character recognition; Convolution; Convolutional neural network (CNN); End to end recognition; Feature extraction; Image detection; Multi oriented; Object recognition; Pipelines; Scene text localization; Text image detection; Text recognition; Text recognition; Text image detection; Multi oriented; Convolutional neural network (CNN); Scene text localization; End to end recognition

•A unified framework is proposed that takes scene text detection and recognition.•New.i.ReLU layer is introduced which can detect text components (even vertical).•New.i.inception layer is introduced which can obtain broadly varying-sized text.•A novel algorithm is introduced for feature extraction (...