This paper presents a YOLOv5-based framework for real-time Indian food classification, aimed at automating food recognition in various applications such as dietary tracking and restaurant management. The model was developed using a dataset comprising 4,298 images across 12 distinct Indian dishes, split into 2,990 training images, 833 validation images, and 475 testing images. The framework was trained over 10, 50 and 100 epochs, achieving an overall accuracy of 85%. The model's classification performance varies across different food items, with precision, recall, and F1-scores analyzed for each class. Notable results include an F1-score of 0.97 for Gulab Jamun and 0.94 for Rasgulla, indicating high reliability in recognizing these items. However, some classes, such as Jalebi and Momos, showed lower F1-scores of 0.65 and 0.71, respectively, suggesting areas for further model refinement. This study highlights the challenges of distinguishing visually similar food items and demonstrates the capability of YOLOv5 in handling real-time classification tasks. The findings suggest that with further optimization, this framework can be effectively deployed in real-world applications, enhancing the accuracy and efficiency of automated food classification systems.