Pizza

Image Based
Calorie Calculation
B.Tech Final Year Project

Under the guidance of

Prof. Kalyani Kadam

Members:
Suyash Phatak - 19070122125
Sayan Sahu - 19070122150
Sharath Pai - 19070122155
Pranhav Vanarote - 19070122190
Contents
• Problem Statement
• Planned Objectives
• Literature Review
• System Architecture
• H/W and S/W specification
• Details about Dataset
• Result Analysis
• Future Scope
• References
Problem
Statement
• Our problem statement is to calculate the
number of calories present in each pizza image
while accounting for the various toppings on it.
• The input for the program would be a set of
pizza images with different toppings for training.
• A robust detection model would be trained to
identify the type of pizza and toppings and try to
estimate the volume of the pizza to find an
approximate number of calories that it would
contain.
• A small sample of unseen images would be
reserved for testing.
Planned
Objectives
• Estimate the calorie from
different types of pizzas and
toppings.
• Minimize error of estimated
calories.
• Website Building
• Model Integration
• Testing & Deployment
 Literature Review

Authors Journal Title Models/Algorithms Datasets Used Performance

Used Metrics
Ghazanfar Latif, ICCTA '20: Proceedings of the 2020 6th International Automatic Fruits Calories Estimation Deep Learning, CNN Fruit-360 Accuracy, Loss
Batool Alsalem, Wejdan Conference on Computer and Technology Applications through
Mubarky, Nazeeruddin Convolutional Neural Networks
Mohammad, Jaafar Alghazo

Kaixi Ruan, Lin Shao https://cs229.stanford.edu/proj2015/151_report.pdf Calorie Estimation From Fast Food SVM, Deep Learning Pittsburgh Fast Accuracy
Images Food Image
Dataset

Yanchao Liang, Jianhua Li https://arxiv.org/abs/1705.07632v3 COMPUTER VISION-BASED FOOD Faster R-CNN, ECUST Food Mean Error
CALORIE ESTIMATION: DATASET, GrabCut Algorithm Dataset
METHOD, AND
EXPERIMENT
 Literature Review

AUTHORS JOURNAL TITLE MODELS/ALGORITHMS USED DATASET USED PERFORMANCE METRICS

LANDU JIANG, BOJIA QIU, XUE LIU, IEEE Access Volume 8 DeepFood: Food Image Analysis Faster R-CNN, VGG-16 FOOD20, FOOD101, mAP, Accuracy
CHENXI HUANG AND KUNHUI LIN and Dietary UEC-FOOD100, UEC-
Assessment via Deep Model FOOD256

Gregorio Villalobos, Rana Almaghrabi, 2012 IEEE International Symposium An Image Processing Approach for SVM Nutrient Info Database Accuracy
Parisa Pouladzadeh, Shervin on Medical Measurements and Calorie Intake Measurement
Shirmohammadi Applications Proceedings

Parisa Pouladzadeh, Abdulsalam Springer, ICIAP 2015 FooDD: Food Detection Dataset for Graph Cut Segmentation, Deep FooDD Dataset Accuracy
Yassine, Shervin Shirmohammadi Calorie Neural Networks
Measurement Using Food Images
System Architecture
• Food recognition is a pre-existing concept whose aim
is to find and identify any food items present in the
given image. Further the goal is to estimate the calorie
of the entire pizza along with the toppings which is
being recognized.

• We use the Mask-RCNN framework for identifying

the food items. For performing calorie calculation, we
use a standardized approach by calculating the values
using the calories per 1 gram of toppings present in
the item.

• The pizza image is provided as an input to the model

and based on the standardized values, the approximate
calorie in the entire pizza is found.
H/W and S/W Specifications
Dataset Details
• We have created our image dataset by
using images.cv website which provides
various image datasets for computer vision
and machine learning applications.
• We have segregated images into three
folders which is training, testing and
validations.
• Training: 2,975, Testing: 889, Validation:
304, Total Images: 7230
• After preprocessing, we have reduced the
number of images to 7000.
Result Analysis

Metric Value
Precision 0.9666
Recall 0.5166
F1-Score 0.6734

Metric Value
mAP 0.9385
mAP 50 0.9124
mAP 75 0.8722
Pie Chart
of
Calories
Future Scope
In the future, this project could be improved where the model would be further enhanced
to work for a broader set of images and be more accurate overall.
References
• https://www.who.int/news-room/fact-sheets/detail/obesity-and-overweight
• https://www.cdc.gov/obesity/data/obesity-and-covid-19.html
• https://arxiv.org/pdf/1705.07632.pdf
• https://arxiv.org/pdf/1811.00982.pdf
• https://www.site.uottawa.ca/~shervin/pubs/FoodRecognitionDataset-MadiMa.pdf
• https://cs229.stanford.edu/proj2015/151_report.pdf
• https://www.calories.info/food/pizza
• https://www.popsugar.com/fitness/Calories-Pizza-Toppings-19209681
• https://www.bacinos.com/pizza-calories-nutrition/
Thank You