Related Experiment Video
Updated: Mar 29, 2026
05:47
Simulation of a Scaled Assembly Process with Collaboration of a Robotic Arm and Monitoring through a Vision System for Quality Control
Published on: August 29, 2025
615
Advanced Sensitive Feature Machine Learning for Aesthetic Evaluation Prediction of Industrial Products
Jinyan Ouyang1, Ziyuan Xi1, Jianning Su1
1School of Architecture and Art Design, Lanzhou University of Technology, Lanzhou 730050, China.
Journal of Imaging
|March 27, 2026
Summary
This study introduces a machine learning framework for objective product aesthetic evaluation using sensitivity-aware features. It enhances prediction accuracy and provides design insights for industrial products, particularly in automotive design.
Area of Science:
- Industrial Design
- Artificial Intelligence
- Machine Learning
Background:
- Product aesthetics significantly influence consumer choice.
- Current aesthetic evaluation methods suffer from subjective biases and AI's black-box nature.
- Quantitative assessment of product aesthetics is challenging.
Purpose of the Study:
- To develop an advanced machine learning framework for quantitative aesthetic evaluation of industrial products.
- To address subjective biases and improve the interpretability of AI in design assessment.
- To provide reverse design insights for product optimization.
Main Methods:
- Developed an aesthetic index system with quantitative formulations for product form.
- Integrated grey relational analysis (GRA), coefficient of variation method (CVM), and TOPSIS for weight determination.
- Proposed a novel model-improved lung performance-based optimization with backpropagation neural network (ILPOBP) using maximin latin hypercube design (MLHD).
- Utilized Shapley additive explanations (SHAP) for model interpretability.
Main Results:
- The ILPOBP model achieved high accuracy in predicting aesthetic ratings from limited morphological data.
- Demonstrated superior performance over baseline models with a mean absolute relative error (MARE) of 4.106%.
- Identified six key indicators through sensitivity analysis for dataset creation.
Conclusions:
- The proposed framework offers a robust and interpretable approach to aesthetic evaluation in industrial design.
- Machine learning can effectively overcome limitations of traditional subjective assessments.
- The study provides valuable tools for optimizing product aesthetics in automotive and other industries.
