Extracting human silhouettes from video for gait recongition

Hello everyone. Recently I have been working on a project related to Gait Recognition. I was trying to implement one of the papers.Paper: GaitSet: Regarding Gait as a Set for Cross-View Gait Recognition . I was looking for an approach where I can extract human silhouettes from the video. I tried some codes from the OpenCV website but the silhouettes I got are noisy. I was wondering if there is any better implementation (like image segmentation). I would appreciate your advice on this topic. Thank you.