In the present technological era, the incorporation of automation is considered crucial for the performance of any organization. Due to technological advancements, complexity has increased, leading to difficulties in processing, and different hardware is utilized for implementation. Consequently, it is imperative to explore feasible solutions to reduce complexity and enhance the applicability of any modus operandi in real time. The purpose of this article is to analyze the potency of deep learning models on the different parameters affecting the processing phenomena to identify the best architecture for creating a real-time food detection system integrated with a Calorie Counter. Based on the literature review, the complexity of different architectures was studied. In this study, the dataset comprises 36 classes, in order to analyze the effect of diverse deep learning structures over a multi-class dataset for elucidating classification. The results of the study show that a lightweight system outperformed a heavyweight system. The findings of the study conclude that MobileNetV2 architecture, which could be utilized as the best real-time multi-class food detection system, is the lightweight architecture system integrated with a Calorie counter. This study will be beneficial for the different stakeholders in the food industry in different aspects such as marketing and development of products, integration of technology, marketing strategies, regulatory compliances, etc. The result depicted in the paper may be generalized in the food industry and trustworthy in nature.