This paper presents a new full-size humanoid robot, "Adam," and introduces a novel whole-body imitation learning framework that enables the robot to achieve human-like locomotion. The robot is designed with a modular, cost-effective structure and high-performance actuators, allowing it to perform complex tasks with human-like characteristics. The framework uses adversarial motion priors to learn from human motion data, significantly improving the robot's adaptability and performance in real-world scenarios. The framework also addresses the Sim2Real gap by incorporating cross-validation and feedback adjustment steps, enabling the robot to adapt to dynamic environments. The experiments demonstrate that Adam can achieve human-comparable performance in complex locomotion tasks, marking the first time human locomotion data has been used for imitation learning in a full-size humanoid robot. The framework is not limited to Adam but can be applied to other humanoid robots. The results show that the proposed method provides a new perspective for future motion learning and optimization in humanoid robotics. The contributions of this paper include the development of a new humanoid robot, the creation of a novel imitation learning framework, and the demonstration of the framework's effectiveness in real-world applications. The framework enables the robot to perform complex motion tasks with high accuracy and adaptability, paving the way for further research and development in humanoid robotics.This paper presents a new full-size humanoid robot, "Adam," and introduces a novel whole-body imitation learning framework that enables the robot to achieve human-like locomotion. The robot is designed with a modular, cost-effective structure and high-performance actuators, allowing it to perform complex tasks with human-like characteristics. The framework uses adversarial motion priors to learn from human motion data, significantly improving the robot's adaptability and performance in real-world scenarios. The framework also addresses the Sim2Real gap by incorporating cross-validation and feedback adjustment steps, enabling the robot to adapt to dynamic environments. The experiments demonstrate that Adam can achieve human-comparable performance in complex locomotion tasks, marking the first time human locomotion data has been used for imitation learning in a full-size humanoid robot. The framework is not limited to Adam but can be applied to other humanoid robots. The results show that the proposed method provides a new perspective for future motion learning and optimization in humanoid robotics. The contributions of this paper include the development of a new humanoid robot, the creation of a novel imitation learning framework, and the demonstration of the framework's effectiveness in real-world applications. The framework enables the robot to perform complex motion tasks with high accuracy and adaptability, paving the way for further research and development in humanoid robotics.