For this assignment, I did a draw-to-order experience with DoodleNet model. The system allows customers to sketch their food items like pizza, hamburgers, donuts, or ice cream on the canvas, and the neural network recognizes the doodles, which add the identified items to the order cart. After submitting the order, there is a digital receipt that shows the food and a back button which returns to the main page. Also the interface has a menu board displaying available food items with emojis.
I ran into an issue where the doodle weren’t recognized accurately. I tried adjusting the stroke weight to 10 instead of 16, and it doesn’t work. Later I changed the value of confidence lower so that it had a higher tolerance rate.
p5 sketch: