Part Identification App

This project was a case of improving upon an existing project. There was a large conveyer belt in the corner of one of the warehouses. The conveyor belt was very useful for the staff as oftentimes items will come into the warehouse untagged with their ID. Staff could place the unknown item on the belt and an AI model would predict the ID, which the staff member could verify and tag. However, the problem with this approach was twofold. The conveyer is very large and had to be lugged out each time an unknown item came in, this took up time, manpower, and space. The conveyor also took substantial time for the item to go down the belt, have images taken, and come back out. This problem is exacerbated by their only being one of these machines, and the potential for hundreds of unknown items.

Noticing potential improvements, I suggested a mobile app connected to the same API as the existing solution. This way a worker could take a top-down photo anywhere in the warehouse speeding up the results over the conveyor solution. This solution also allowed multiple people to identify items at once time. For a rapid prototype I made the app in Godot. I created a barebones UI that let the user take a photo or upload a photo from their own devices, then see the results of the model.

This solution vastly improved upon the old system speeding the recognition of unknown items as described, freeing up time, and manpower. Unfortunately, a better version of the app could not be developed due to the AR4 project which you can also see on this site.