In the realm of culinary culture, classifying desserts presents an interesting challenge. Due to the complexity as well as the variety of dessert types across various regions. This study present a comprehensive dataset of dessert images that have been specially selected for Bangladeshi dessert classification. In this research, there is a thorough collection of high-quality pictures showing the vast range of traditional sweets from Bangladesh. The collection comprises a varied assortment of traditional Bangladeshi desserts, demonstrating the richness and diversity of the countryʼs culinary traditions. This study seeks to develop reliable disease recognition models through the inspection of different image processing techniques and deep learning algorithms such as MobileNet. Specifically, these models' performance is subjected to a rigorous evaluation process using a set of standardized evaluation metrics to achieve a 98 % overall test accuracy. This paper presents a valuable resource for researchers and practitioners interested in the domain of culinary image classification, particularly focusing on Bangladeshi desserts. The availability of such a dataset will foster advancements in machine learning techniques tailored to culinary applications, ultimately enriching culinary heritage preservation and promoting cultural diversity through technology.