A picture of Shelfie from the side.
Shelfie's desktop client. The client has separated and labelled each item it has seen based on a custom collection we trained on Indico.
Shelfie is an automated storage solution aimed to be low cost and simple. We wanted to make a solution similar to what is found in Amazon warehouses but for smaller businesses and home use.
What it does
Shelfie is built of two systems: its inventory detection system and its robotic retrieval system. Shelfie has a camera which looks at a shelf on which items are placed. It then tries identify objects placed on the shelf. If Shelfie mislabels an item, the user can retrain Shelfie to label the item correctly. After labeling, Shelfie can be told to go to an item and take it off the shelf.
How we built it
Shelfie is both a desktop client and a physical robot. The desktop client is a C# Windows Form which executes some Python scripts. The Python scripts handle scanning the shelf, object detection, retraining its object detection model with user input, and controlling the robot. The object detection is done with a custom collection on the Indico API. The robot was made with the Particle Photon board, a solenoid, continuous servos, and toy trains (to act as wheels).
What's next for Shelfie
Improving how Shelfie takes items off shelves