During the past year and a half, the world has witnessed a raging pandemic that has forced countries to implement lockdowns in order to combat the spread of Covid-19. Taking precautions meant, people were confined to their homes for the most part which has resulted in them hesitating to visit hospitals for regular health checkups. During the pandemic, our immunity is of utmost importance and signs of health deficiency should not be ignored. But having said that people hesitate to visit hospitals to conserve time and for their safety, a tool to facilitate these regular health checkups was essential. Our concept is to develop a smart application that people can use to identify any health deficiencies that they might potentially have by taking photos of your body parts. The application will utilize a pre-trained artificial intelligence model (Convolutional Neural Network) to analyze the photos and identify certain characteristic features to diagnose any potential health deficiency. If any health deficiency is identified, the smart app will also provide appropriate diets to increase the amount of a particular vitamin/mineral in the body to the optimum level.

Technologies used

The dataset was first collected with the help of python scripts from google images. Furthermore, the Convolutional Neural Network (CNN) was developed with the help of MATLAB in which the dataset was trained and tested in it as well. The image recognition and feature extraction was done with the help of AlexNet. Furthermore, the prototype of the Mobile App was built with the help of Android Studio. Finally, the video for this presentation was made and compiled with the help of Microsoft PowerPoint and Adobe Premier Pro.


One of the most difficult aspect in this project was the data collection for making the dataset. This is due to the fact that CNN model requires the dataset to be large and needs to have optimal type of images. This could only be done by properly analyzing and cautiously trimming out any irrelevant images that were present in the dataset that could potentially lead to inaccuracies in the model. Another difficult aspect was the the classification of different problems that leads to different deficiencies. This is because the datasets need to be optimal and requires any irrelevant images that were present to be removed to make the model accurate.


We are proud to say that our model had an overall accuracy 89.42% which was done on the weekend. This shows that our model is able to accurately diagnose different problems which leads to different potential nutrient deficiencies. Our model also had an 100% accuracy in identifying whether there was a deficiency to the user or not. This shows that our model can very accurately analyze and diagnose for deficiency for users.


We have learned all the different types of neural network and for this project how to implement a Conventional Neural Network to diagnose a person deficiency or disorder by scanning their lips, eyes, mouth and nails and providing ways on how to rectifies them. We have learned on how to harvest large amount of data required for our project with the help of python and downloading them all at one instead of the downloading each of them individually which is a time-consuming process and on how to sort them by remove the unwanted images that will affect the result. We learned how to analyze our data and diagnose it using MATLAB and to design our app. We have used Android Studio for designing our app and the prototype as well. This project helped us in diversifying our skills even further.

Next steps

Our next step is to further increase the accuracy of our model to around 95% which can be done by increasing more amount of data in our dataset and further optimizing our dataset as well. We also want to publish our App in the Play Store when it is completely done as well.

One-minute video