Passed out in aisle 5

Aisle corner detection.
Passed out (Getty Images)

Inspiration

We are under time pressure. Going shopping now, but later we have a date we cannot miss. Our list is long and we have to figure out the fastest way through the store.

What it does

The algorithm finds the fastest way between articles on our shopping list in our local Walmart supermarket.

How did we do it

We are first starting a Traveling Salesman Problem (TSP) disregarding the shelves to establish a rough order of target locations to be visited. Then we run an A* search between two target locations and continue from one to another until we have visited all target articles. The A* can also take into account the business of aisles and the least problematic path between targets. A more detailed description is in the Jupyter notebook in the GitHub Repository.

Challenges I ran into

Everything. We tried corner detection on the shelves in hopes that it would be less computationally demanding than A*. In the end we decided for A* even if it takes longer than expected. Using the TSP first helped avoiding a large number of target combinations and expedited the process. This way no clustering step for *A searches had to be implemented.

Accomplishments that I'm proud of

We tried.

What I learned

You need to try out at least three different image processing libraries until you find one that can resize images. Control-click to type in multiple locations at the same time.