Have you ever gone to the grocery store and seen your favorite item not there on the shelves? You come back the next day, and the day after, but it is still not there? Our group believed that the retail industry has been struggling in terms of inventory management, a study by Shiau Wei Chan from the Universiti Tun Hussein Onn Malaysia stated that discrepancies in the inventory have a mean score of over 3 when it comes to inventory management.

What it does

Flyview Inventory Tracker(F.I.T.) helps retail workers take inventory and manage their supplies in a more efficient and safe manner. The web app that we created comes with a Python backend which allows store owners and warehouse workers to scan their floors alongside their shelving units and items present on the shelves. All it takes is a press of a button or a voice command, and the drone can be sent to do our tasks, like finding how many water bottles are on the shelves.

How we built it

Our drone uses Python and a few libraries which give the drone access to voice recognition, involving natural language processing, and object detection using OpenCV. Alongside our drone, we built a state-of-the-art machine-learning model which utilizes convolutional neural networks, with Tensorflow and allows our drone to detect and identify how many of one object there are in a certain frame.

Challenges we ran into

While training our model, we weren't able to widespread data on the inventory object we were training for. To solve this, we implemented some techniques such as using

Accomplishments that we're proud of

We built a program that allow us to control the drone with voices, train a CNN model with our own dataset, a website that play the interface. We are happy that we have done so much work in 24 hours

What we learned

This hackathon give us a lesson of importance of datasets. By participating the HackAI, we first time try to understand CNN and built our first customized dataset (around 100 images taken by our phone and drone camera)

What's next for Flyview Inventory Tracker

For next few steps, we want to increase our data pool, so the drone can recognize more items besides water bottles. Also we try to add features like Logical Programming, Path mapping and Multi-modal model that can match object to their location


Chan, Shiau Wei. (2017). Factors Influencing the Effectiveness of Inventory Management in Manufacturing SMEs. IOP Conference Series: Materials Science and Engineering. 226. 10.1088/1757-899X/226/1/012024.

Built With

Share this project: