This paper explores the use of deep learning methods for human gait recognition, focusing on the application of CNN and VGG19 models in conjunction with GEIs to classify and extract gait features. We used the CASIA (A & B) datasets to evaluate performance of the models under various conditions, including normal walking, walking while wearing a coat and walking while carrying a bag. The paper provides a comprehensive evaluation of the CNN and VGG19 architectures, comparing their performance based on accuracy. Through experimentation, the CNN model achieved a testing accuracy of 97.93%, offering a lightweight and efficient solution for gait recognition, while the VGG19 model, leveraging transfer learning, achieved a training accuracy of 96.93%. We also applied K-fold cross-validation to assess the generalizability of the models. Furthermore, we examined the impact of training the models with different gait conditions and the challenges presented by hidden abnormalities. This paper highlights the advantages of using CNN for efficient gait recognition and discusses the implications of our findings for further research in gait recognition systems.