Deep Learning Approach for Sign Language Recognition Using DenseNet201 with Transfer Learning
Yasir Altaf, Abdul Wahid, Mudasir Manzoor Kirmani
Abstract
Transfer learning has been utilized to solve many complex real-world problems. Over the last several years, transfer learning had many applications in image and video recognition. To get higher recognition rates, deep and wider architectures of the convolutional neural networks (CNN) have been designed. In this research, we proposed a novel transfer learning-based model using a popular CNN architecture called DenseNet201 for the recognition of Indian Sign Language (ISL) hand gestures. We applied transfer learning to DenseNet201 by freezing some of its layers to retain its knowledge of generalization and fine-tuning the remaining layers for ISL dataset. Pre-trained DenseNet201 was used to extract the features of the gesture images. To classify the ISL gesture, custom layers were added to the pretrained DenseNet201 model. The proposed model helped to achieve higher accuracy of 100%.