Member 1 : Ayush Agrawal, 2017A8PS0604G, f20170604@goa.bits-pilani.ac.in, Tensorflow model Development
Member 2 : Mayank Harsani, 2015A7PS0010G, f20150010@goa.bits-pilani.ac.in, Application UI and Backend Development
Every day seeing the mess food wastage amount rise up made us think, if a small community of \ people produces 100 kgs of food wastage per day, How much total amount of food waste is \ produced globally? A 2012 report from the National Resources Defense Council (NRDC) revealed \ a list of statistics that show just how big the food waste problem is in America:- \ ● Americans throw away $165 billion of food each year. \ ● 40% of food is wasted in the United States every year. \ ● 20% of food that the average American buys is never eaten. \ All of this wasted food is a huge contributor to climate change. It takes a ton of resource to \ produce food. On top of that, you have all of the energy it takes to keep it cold and transport it \ around the country. And when food decomposes, it isn’t just stinky. It releases potent \ greenhouse gases. Basically, we’re trashing our planet to grow more food that no one eats. But \ here’s the thing : No one actually likes wasting food. It just something we haven’t been paying \ much attention to. Of all the challenges in the world, reducing the amount of food is relatively one \ of the easiest.
Consumers collectively make up the largest portion of food waste. And you don’t have to be an \ expert too understand why food is going to waste in our homes. We’re all busy on the go and we \ buy food without thinking if we really need it. There’s even been a little bit of research to show \ that once something goes in the refrigerator, it’s actually worthless to us than before. A lot of the \ waste in our society happens when we come down to the choice and wanting to have the option \ to eat something at any time whether or not we use. We store a lot of food inside our refrigerator \ but that’s the problem : we store our food and then forget about it. Unfortunately, many of us do it \ even when we don’t intend to. If you are not keeping track of the expiration date of your foods, \ you are going to end up wasting food.
● Problem Statement :- \ “I find it difficult to keep track of the food that I have and hate to throw away expired food.” \ To prevent food wastage due to expiration, develop an application that tells you when \ your food is going to expire so that you can use it before it does. It should intelligently \ send you an alert when your food is close to expiring. \ ● Deliverables :- \ Android based application to prevent food wastage that lets you scan the item or the \ barcode and automatically figures out the name of the item, classifies it into the type of \ food, lets you record the shelf life of your goods and reminds you to use them before they \ expire. It will estimate an expiration date and to make things a lot easier and will also \ keep a track of the previous items you’ve scanned so that you don’t have to scan again.
There has been significant work done on the Image Classification Subsystem of this application \ and some applications based on basic implementation of this idea have already been developed. \ Here, we are listing some of the work we found that matches our project : \ ● Keeping Track of Expiry Dates & Reducing Food Wastage \ ● Fridgely - Food Expiration Date Tracker & FreshFridge \ Every project/work that we found depends on addition of food item manually or through barcode \ scanning but none among them scans the food item.
Development : Desktop/Laptop \ ● NVIDIA® GPU card with CUDA® Compute Capability 3.5 or higher enabled for Tensorflow. \ NVIDIA® GTX 1070 or above preferred. \ ● Quad core Intel Core i7 Skylake or higher \ ● 8Gb RAM recommended \ ● 16Gb of disk space \ ● 1280 x 800 minimum screen resolution \ Tools:- \ ● Android Studio IDE 3.3.1 \ ● Anaconda/Jupyter Notebook \ ● Git \ ● Flutter v1.0 \ ● Firebase SDK v16.x.x \ Android Application Requirements : \ ● Android 5.0 ( API 21) or higher is required.
Design GUI of the application \ ● Finalize a simple user friendly design for the complete application. \ Allow adding products through barcode scanning \ ● In the first iteration, allow user to add and categorize food through barcode. \ Data Cleaning \ ● Study and familiarize ourselves with CNN, Keras, Tensorflow and TensorFlow Lite . \ ● Clean, Refine and Optimize the datasets for food classification training: Fruits , Vegetables , \ Other Groceries . \ Design a neural network for automatic identification \ ● Finally, training a convolutional Neural Network for classifying food images using \ appropriate framework. \ Converting the TensorFlow model \ ● Freezing the graph and optimizing the model file for mobile apps. \ ● Export the project to TensorFlow Lite in order to achieve low latency and hardware \ acceleration. \ Integration of Model in the mobile application \ ● Integration of converted .tflite model with the mobile application, which will now support \ adding items both through image identification as well as barcode scanning. Scanning the \ item will display its name, classification and estimated expiration date. \ Refactoring and Testing \ ● Refactoring the code by following good development practices. \ ● Developing test suite which involves everything from Unit, Integration and End to End \ testing. \ Documentation \ ● Well organized documentation for Developers as well as for the Users. \ A buffer of two weeks has been kept for any unpredictable delay.
● How to export Tensorflow and Keras Model to ML Kit. \ ● How to export Tensorflow Models to Tensorflow Lite for portable devices. \ ● Research Papers : \ ○ http://vim.ustc.edu.cn/?product=vegfru \ ○ https://www.researchgate.net/publication/321475443_Fruit_recognition_from_images_using_deep_learning \ ○ http://cs231n.stanford.edu/reports/2017/pdfs/607.pdf \ ○ http://ijabe.org/index.php/ijabe/article/view/2690/pdf \