Challenges related to gender diversity and the aging European workforce can be addressed through the adoption of new technologies aimed at enhancing production performance. With the combination of sustainability and new technologies, Industry 5.0 introduces a new manufacturing paradigm that emphasizes human-robot collaboration to enhance productivity, flexibility, and worker well-being—particularly beneficial for Small and Medium-sized Enterprises (SMEs). In this context, this paper focuses on a vision-based Human-Machine Interface (HMI) enabling intuitive, gesture-driven interaction between a human operator and a collaborative robot (UR5e) during shared assembly tasks. AI techniques (neural networks, deep learning) are exploited through Google's Mediapipe for hand gesture recognition and the Robot Operating System (ROS) for real-time robot control. A custom dataset has been created to train the gesture recognition module. A simulation has been elaborated in the Gazebo environment, to demonstrate how the operator can naturally instruct the robot to halt or proceed with tasks, fostering collaborative execution on a shared workpiece. The proposed approach offers a lightweight, low-cost solution for SMEs seeking to implement flexible human-robot collaboration with minimal hardware and setup.