Comparing CNN Architecture for Indonesian Speciality Cuisine Classification

Abstract

Indonesia's diverse and flavorful cuisine is a hidden gem, reflecting the nation's rich history and cultural tapestry. However, many of these culinary treasures remain undiscovered by a wider audience despite the popularity of beef rendang. This study represents a fascinating blend of technology and gastronomy, using smart computers to unravel the secrets of Indonesian flavors. This research employs one of the most popular neural networks methods called Convolutional Neural Network (CNN) to shine a light on many citizens' favorite regional specialty cuisine which is Padang cuisine from West Sumatra, Indonesia. Gathering a collection of 993 images from 9 various dishes, the machine is trained to automatically recognize these unique culinary delights. Among several different Convolutional Neural Network models trained and tested, DenseNet-201 emerged as the top performer, showcasing remarkable accuracy, precision, recall and f1-score higher than 0.90. By harnessing the power of advanced neural networks, we not only gain insights into the intricacies of the region's culinary traditions but also pave the way for a deeper appreciation and understanding of the cultural significance embedded in every bite. Beyond this research technological achievements, it also emphasizes the importance of preserving and promoting Indonesia's diverse culinary heritage and rich tapestry of global food heritage.