Given its uses in access control, security, and surveillance, computer vision research on gait recognition has grown significantly in recent years. In order to extract silhouettes for identification, gait recognition algorithms have historically depended on the presumption that people are walking perpendicular to the camera’s axis. Nevertheless, this constraint limits the adaptability and practicality of these systems, particularly in real-world situations where individuals might not consistently walk in such regulated spaces. Because of this, current research has concentrated on tackling the problem of identifying gait patterns independent of walking direction. This entails creating methods and algorithms that can recognize people precisely even when they aren’t moving parallel to the camera axis. These kinds of developments are essential for improving gait recognition systems’ usefulness and efficacy in a range of real-world scenarios. Removing the restriction on walking direction broadens the use of gait recognition but also presents new difficulties, such as adjusting to changes in walking angles, occlusions, and surrounding conditions. To increase the resilience and accuracy of gait recognition systems in less-than-ideal circumstances, researchers are investigating novel strategies such as deep learning techniques, multi-view analysis, and the fusion of several modalities (such as depth information).