Soybeans tempeh image dataset for tempeh maturity detection.

Area of Science:

• Food Science and Technology
• Agricultural Engineering
• Computer Science (Machine Learning)

Background:

• Tempeh, a traditional Indonesian fermented soya bean product, is a vital animal protein substitute globally.
• Current tempeh production relies on manual observation throughout its 1-4 day fermentation process.
• The manual process is labor-intensive and presents opportunities for automation to improve efficiency.

Purpose of the Study:

• To explore the potential of computer image detection with machine learning for automating the observation of soya bean fermentation in tempeh production.
• To develop an image dataset capturing the key stages of traditional tempeh fermentation.
• To enhance working time efficiency in tempeh manufacturing through technological integration.

Main Methods:

• Development of an image dataset focused on the soya bean fermentation process for traditional tempeh.
• Image data acquisition using smartphone cameras at three distinct fermentation stages: Day-0, Day-1, and Day-2.
• Data preprocessing included image filtration and cropping, with images organized into time-based folders.

Main Results:

• A curated image dataset representing the visual progression of soya bean fermentation for tempeh.
• Distinct visual characteristics identified for each fermentation stage (Day-0, Day-1, Day-2).
• Demonstrated potential for machine learning-based image analysis to monitor fermentation.

Conclusions:

• Computer image detection and machine learning offer a viable solution for automating tempeh fermentation monitoring.
• The created image dataset serves as a foundation for developing automated quality control systems.
• Automation can significantly increase efficiency and consistency in traditional tempeh production.