Abstract
Tempeh is a traditional Indonesian food made from soya beans, but it has recently been improved by using other types of beans as the base source. Tempeh has been consumed around the world for a long time and is one of the food substitutes for animal protein. The creation of tempeh until now still utilizes the traditional method from the beginning, observation, until the collection of the final product, which still includes the human role inside the process. The process of creating tempeh products takes approximately 1 to 4 days. The observation of tempeh creation in the soya bean's fermentation phase has automation potential by implementing computer image detection with a Machine Learning method that can increase the working time efficiency. Based on the potential, this research forms an image dataset containing the soya bean fermentation process for creating traditional tempeh products. The image data were taken by capturing product photos directly using the smartphone camera. There are 3 stages at the time of tempeh product creation for image data collection. Starting from the zero-day process, the soya beans that have been through the cooking and washing process and have been mixed with tempeh yeast (Rhizopus sp), are put into the product mold package. The first-day process is when the wrapped soya beans are at the fermentation stage and turn into the tempeh products and can be identified by warm water dew and a little tempeh mold. On the second-day process, the soya beans have been fully covered by tempeh mold and have been solidified. The data created from this research is in the raw image data that has been through the process of filtration and image cropping. The image data is grouped by folders named by the time of the fermentation process to form the tempeh product, namely Day-0, Day-1, and Day-2.