Table 3. Comparative analysis of interactive editing, optimization algorithms, and deep learning-based scene synthesis researches.

Items	Interactive Editing	Optimization algorithm	Deep-Learning
Methodology	User input-based, object recommendation and group editing	Definition of cost function, mathematical model such as global optimization, sampling, etc.	Training large datasets
Data dependency	Object model resources	Sample scene, user-defined parameters	Large-scale 3D scene dataset
Cost of calculation	Real-time interactivity	Complex computational complexity due to nonlinear optimization	High cost in learning process, relatively fast inference process
Degree of user input	High, user-dependent	Low, initial constraints and parameter settings	Usually, parameter setting during learning process
Diversity of results	Depends on user's capabilities	Depends on the design level of the constraints	Depends on the dataset
Strength	Intuitive, real-time editing capabilities	Setting explicit constraints	Automation, realism, diversity
Weakness	Highly dependent on user skill	High computational volume and difficulty in designing cost functions	Depends on dataset quality and availability, and has high training costs.