We do not use AI (deep learning or machine learning) for image analysis.
Our company adopts a neuroscience-based approach because of its advantages in "interpretability" and "versatility." We do not utilize AI-driven statistical approaches.
| Image Analysis with Deep Learning | EXplainable KANSEI | |
|---|---|---|
| Processes images statistically | Implements mechanisms based on neuroscience | |
| Interpretability | Difficult to interpret why certain results occur, making it challenging to derive improvement actions. | For example, mechanisms such as brightness, color, shape, or faces contributing to attention are clearly defined, making the process understandable and easier to translate into actionable improvements. |
| Versatility | For example, if trained with web images, it cannot be used to evaluate shelf or package images. | Since it mimics brain processing mechanisms, it can generally handle any type of image. |
* Deep learning or machine learning is introduced in cases where interpretability can be ensured, such as text extraction, face recognition, or text analysis.