Bi-LSTM Model based Person Gender classification using Gait Energy Image and Transfer Learning Techniques

Several applications have found that human gait provides important motion information. Irises, thumbprints, faces, gait, voice, and many other characteristics are utilized by human recognition AI to identify individuals in a distinctive way. The ever-increasing amount of online image data has created a significant need for Artificial Intelligence (AI) algorithms capable of executing classification tasks such as Gender classification. The emergence of Deep-learning techniques such as Convolutional-networks and RNN has resulted in a swift increase in correctness in certain picture categorization operations. In this work, a unique approach using LSTM is proposed to predict the gender of an individual by using many pre-trained transfer learning models. With the purpose of evaluate the effectiveness of this scheme, we conduct training on a Bi-LSTM model using an OU-ISIR-MVLP (Multi-view large population) Gait Dataset. The dataset comprises 10300 participants, balanced between males and females, and includes photos captured from 14 various viewing angles. Objectives to determine the age and gender of users. The top level of accuracy achieved by the suggested work in gender classification is 89%.