Go click pic and eat
GetSetFit application is aimed to assist the people with regard to their fitness, daily diet, monitoring the intake of calories, proteins, carbohydrates, fat and fibre.
Just on single click user can upload the the image of his diet and our application will use machine learning model to to find the item and will prepare the report of all calories, proteins,carbohydrates, fat and fibre intake and will also tell the required amount by person in accordance with his or her BMI.
Link for model checkpoints: https://drive.google.com/drive/folders/1hROz-dhJrXPrl8bAnqthIledLCRrXgUp?usp=sharing
We had trained the multi-class Classifier using Convolution Neural Networks over 7 broad categories: Roti, Bread, Yellow-daal, Rice, Boiled-Egg, omlet, chai.
further on getting more time we could scale it at as many items as we want and could give sufficient time to train our model.
What it does is :- In this application user is just required to scan their food and upload it . After that the trained machine learning model will predict the food item and will provide you the calories,protein ,carbohydrates,fats,fibre of that food item and will maintain your chart for whole day.
Our app at starting calculates BMI of person and sets the initial status for the person just by using simple universal logics. like:
If BMI > 18.5 and BMI <= 24.9
Status =Normal
Solution = MaintenanceBMI -- weight/(height*height)
Men_BMR = 66.4730 + (13.7516 x weight in kg) + (5.0033 x height in cm) – (6.7550 x age in years)
Women_BMR = 655.0955 + (9.5634 x weight in kg) + (1.8496 x height in cm) – (4.6756 x age in years)- Learning Rate: 0.0001
- Optimizer: Adam
- Loss function: sparse_categorical_crossentropy
- Activation function: Relu, Softmax
- Layers used are: Dense, Dropout, GlobalAveragePooling2D, Flatten, Conv2D, MaxPooling2D
In future we can train the machine learning model with more number of images so that the prediction accuracy of the model improves. Currently we can classify the foods in 7 categories but later on we can use more data and computational power to train a better model which can classify foods into various other categories. User can have his own dieting history in the profile panel which can help him judge his own past performace. With the help of proper of dietician we can have three major diet plans of reducing weight , gaining weight and mainting the current phsique in the future.
This website/software can generate revenue through these models.
We can make user pay for the service. Standalone user can register in our application to track their dieting activities. They will be provided with zero advertisement experience and they can choose from the available dieting plans we have or else can simply use our application for tracking and judging their calorie intake. People with obese condition, or people who are looking to gain weight and the gym freaks or physically fit people will be the target group for this category.
The user will be given the same priviliges as in model 01 but the advertisement will be shown to this category user. User in this category will also be a standalone user and he will be shown with the advertisement of health cauntious brands, gym brands, dieticians in the location near him and also the restaurants who offer diet food.
The software or website will be given as a service to various dieticians to track their patients growth and improvement. Dietician can look at the consumed diet of their patients regularly without physically meeting them. Dieticians can provide their diet plans or suggestion to particular patients through our portal. We will use django-tenant-model to register every dietician as a tenant and every tenant will have his/her private patients. This dieticians will be asked to pay for our service.