Real- Time Image-based Calorie Estimation using YOLOv4 Object Detection

This research study presents an innovative approach to food calorie estimation using deep learning and computer vision techniques. With the growing concern over diet-related health issues, the system offers a convenient solution for individuals to assess the nutritional value of their meals using smartphones or desktop computers. Key components of the system include image acquisition, object detection, segmentation, volume estimation, and calorie estimation. Users capture photos of their meals with standard reference objects, which are then processed using object detection algorithms like YOLO- V 4 Darknet to identify and localize food items. The proposed and widely used YOLO algorithm for object detection operates by first grid-dividing the input picture and then using the grid cells to estimate the bounding boxes and class probabilities. Subsequent segmentation techniques isolate the food items from the background, followed by volume estimation using geometric formulas. Calorie estimation is performed based on the estimated volume and density of the food items using pre-defined tables. Experimental results demonstrate the system's accuracy in estimating calorie content, making it a valuable tool for individuals aiming to make informed dietary choices and maintain a healthy lifestyle.