BetterShopper

Are you paranoid of constantly being scammed? Snatched from deals? BetterShopper allows you to shop safer, and aims to provide a better shopping experience for everyone.

Comment 3

Inspiration

Our inspiration in creating BetterShopper comes from the everyday online scams that happen around the world. More than 60% of Americans say they’ve been a victim of an online scam. Shopping is supposed to be something enjoyable, to re-invest, to have a friendly environment to indulge yourself in. To address this issue, we will be utilizing the power of AI, machine learning, to address this problem.

What it does

BetterShopper uses AI to recognize products and searches it through an algorithm to find products with fair prices to help everyday consumers not get scammed.

How we built it

We used React.JS, Javascript, Tensorflow.JS, Mobilenet pretrained model, and the basic web development tools such as HTML and CSS.

Later, we used reactJs to design the interactive web and scraping. What is web scraping? It's to scrape data off the web, and it will use this data off the web, to build and organize our own.

To use this, we used the API functions, and decided that it would be best to find used products, and lowest prices with high review rate. While this is a two day project, we are looking forward to creating an ML model that will use to sort which reviews are real or not.

To address this problem more in depth, we have created an image classifier, trained with mobilenet, a pre trained CNN model trained on one of the worlds, biggest datasets.. Image-net. So what is a CNN model. CNN stands for (edits) Convolutional Neural Network.

Challenges we ran into

We originally wanted to use a custom trained model that are able to determine specific models and colors, but we ran into issues with React and Tensorflow.js not being able to load the models. For example, loading in a custom CNN model failed to load even after multiple attempts of h5 conversion.

Accomplishments that we're proud of

Using Machine Learning to search fairly priced used items and display them to users. Build a minimalistic and dynamically pleasing using Reactjs which is one of the top growing frame works of web development. Implementing pretrained model alongside with creating our own custom model helped us gain more experience in the world of Machine Learning.

What we learned

We learned how to integrate Tensorflow.js and React.js together, as well as more about web developing and machine learning. As well as learning multiple variations of webscraping, and how to utilize API keys using git to work together. Management was also used in the process, managing our time and our limited resources we had in two days!

What's next for BetterShopper

We plan to continue building it and refine the UI, models to have more accurate results and specific product model numbers. Creating custom models and deploying them to produce better results than the current pretrained one(ensembling).

Extra context

With over 60% of people in America alone, being scammed through online is a problem we need to address. In order for everyone to have a safe environment when supposedly doing this joyous activity, mustn't fall down in the hands of scammers, and bad prices.

Built With

Share this project:

Updates