TechTrade

Inspiration

People are keeping a lot of properties that they don’t use anymore. Digital devices and vehicles are among the most common types of belongings that no longer fit with the owners’ expectations. Unlike clothes, books, bags and other items, which can be donated to charity or exchanged easily, digital devices and vehicles are too expensive and highly electrical to be given away. Therefore, we take them as our target trading products. Normally, people will go to Second-hand groups on Facebook, which supports the fact that there has not yet been a platform designed specifically for trading second-hand items.

What it does

TechTrade is a platform responsible for trading second-hand digital devices and vehicles. Users who logged into their own accounts are allowed to sell and purchase the products at the same time and we also categorize different types of products so that they can easily choose one that meets their demand. Regarding the purchase part, we also provide specific information of the products and some reviews from the other users so as to assist the customers with deciding the most suitable item. For the sellers, we also create an opportunity for them to upload their second-hand products on the homepage with the given price range from our software.

How we built it

The product is divided into two parts: Machine Learning and Web development. We use Python, datasets from Kaggle, and sklearn library for machine learning to train a traditional model. This model is used to predict the car price range using text input (model name, original price, etc...). Then, we aim to use the Convolution Neural Network to detect the product, classify its model, and give the output of that product’s appearance status. However, this CNN model takes much time to be thoroughly trained and has high accuracy. Therefore, we still want to develop our model soon. We use ReactJs to build our website for website development and deploy our machine learning model using Flask and RESTful API.

Challenges we ran into

There were also several issues arising during our making process. First and foremost, at the beginning, we were really confused between different ideas and we had a long discussion on deciding the final topic. After numerous meetings, we finally figured out the best idea and tried our best to finalize the content in order to start building the website. Secondly, it is completely difficult and time-consuming to train a highly accurate CNN model. Furthermore, the website design poses different problems that took us a noticeably long time to finish. The fourth challenge is that we also faced many difficulties while deploying our machine learning model into our website.

Accomplishments that we're proud of

Successfully built a trading platform where people can exchange devices and vehicles, selling their old stuff and buying second-hand ones. We're glad to help to resolve a common issue in assessing the quality of old products. Besides, TechTrade offers tech-users an efficient solution to treat old devices, which supports environmental conservation and saves energy.

What we learned

Thanks to the challenges we suffered during the building process, we’ve learned many things: How to train the CNN model and acquire a high level of accuracy. How to detect the product by using the Convolution Neural Network, classifying its model, and giving the output of that product’s appearance status. How to display the website for the users (both seller and buyer) to reach their needs easily.

What's next for TechTrade

Use AI to conduct a more accurate assessment of the condition of products. Update the assortment of categories in order to serve the customers’ needs more directly.