Episode 1: AI Image Analysis on Human Body
Episode 2: (Coming Soon)
Episode 3: (Coming Soon)
AI Image Analysis Series ①: Can Human Body Be Explained by Numbers?
How accurately can AI analyze a human's body proportions?
This article documents the process of building a custom tool and running live experiments.
The idea of quantifying human body shapes is no longer a matter of curiosity. With advancements in AI and open tools, this form of analysis is now available to anyone. After confirming the feasibility of image-based AI analysis through ChatGPT, I conducted various tests, which revealed both possibilities and limitations. This series presents those findings.
This isn’t about subjective impressions like “pretty” or “looks tall.” It’s about actual measurements—percentages, lengths, ratios—and the methods behind them. Let’s begin with the core of this project: “Image-based Human Proportion Analysis.”
🔍 AI Image Analysis – What Can Be Done?
According to ChatGPT, AI image analysis allows the following tasks:
| Category | Tasks | Examples |
|---|---|---|
| Measurement & Coordinate Extraction | Calculate distances, positions, and ratios from images | Height estimation, body ratio, building size estimation |
| Text Extraction (OCR) | Extracting text from image | Scanning documents, receipts, signs |
| Chart/Graph Interpretation | Recognizing and summarizing data visualizations | Summarize research charts, read bar graphs |
| Layout & Design Analysis | Analyze composition, colors, balance | Design feedback, thumbnail evaluation |
| Image Editing Instructions | Remove/add/change elements and styles | Remove backgrounds, change subjects |
| AI-Generated Illustrations | Create scenes or diagrams on request | Blog illustrations, technical diagrams |
🌐 Global Research Trends in AI-Based Human Image Analysis
| Type | AI Technique | Key References |
|---|---|---|
| Facial Golden Ratio Analysis | CNN, statistical analysis | Prokopakis et al. (2013), Pallett et al. (2010) |
| Facial Attractiveness Prediction | Deep learning visualization | Liu et al. (2019) |
| Height Estimation | Ratio + Deep learning | Günel et al. (2018) |
| Pose Estimation | 2D/3D CNN-based models | Zheng et al. Review (2020) |
Tips:
- Golden ratio ≠ universal standard → mixed research conclusions
- AI excels at automatic learning from data
- Height estimation shows practical potential within error margins
- Body analysis relies on pose estimation techniques
References:
Cureus (2024)
Günel et al. (CVPR 2018)
Liu et al. (2019)
🧪 Tools Used in This Series
While exploring the boundary between what AI can do and what still requires human verification, I created two image analysis tools—despite having no coding experience—using ChatGPT's help.
These tools follow this sequence: Click coordinates → Automatic calculation → Analysis → Final evaluation. They are all browser-based and rely on basic HTML + JavaScript + GitHub Pages.
댓글
댓글 쓰기