A survey on model based approaches for 2D and 3D visual human pose recovery

Area of Science:

• Computer Vision
• Human Pose Recovery

Background:

• Human Pose Recovery research spans 40 years with advancements in data representation and model design.
• Significant challenges persist in achieving accurate human pose recovery in uncontrolled environments.

Purpose of the Study:

• To define a general taxonomy for model-based Human Pose Recovery approaches.
• To methodologically compare current State-of-the-Art (SoA) methods based on the proposed taxonomy.

Main Methods:

• Development of a five-module taxonomy: appearance, viewpoint, spatial relations, temporal consistency, and behavior.
• Comparative analysis of existing SoA approaches against the established taxonomy.

Main Results:

• Identification of key advantages and disadvantages across reviewed human pose recovery literature.
• Structured evaluation of different modeling strategies for human pose recovery.

Conclusions:

• The proposed taxonomy provides a framework for understanding and evaluating human pose recovery methods.
• Further research is needed to address the limitations of current approaches in complex, uncontrolled settings.