Eps 46: Pizza & Computer Vision
— The too lazy to register an account podcast
Pizza quality evaluation using computer vision--part 1 : Pizza base and sauce spread
In the current research the use of computer vision for inspection of pizza base and tomato sauce spread quality was investigated.
Twenty pizza bases were analysed for base area, spatial ratio I (SRI), spatial ratio II (SRII), and circularity.
| Seed data: | Link 1, Link 2 |
|---|---|
| Host image: | StyleGAN neural net |
| Content creation: | GPT-3.5, |
Host
Theresa Barnes
Podcast Content
MIT and QCRI have created a neural network that can look at a picture of a pizza, determine the type and distribution of ingredients, and determine the correct order in which the pizza is stacked for cooking. But that doesn't make it any more precise than a list of instructions on what to eat from your pizza.
As much as the AI understands something, it understands what a pizza should look like from start to finish. The AI has developed a way to visualise what the pizza 'should look like' based on the added or removed ingredients. For example, you can show him a picture of your pizza at work and then ask him to remove the mushrooms and onions. It will generate a picture of the modified cake, but you can then request that the mushroom and onion be removed, or the tomatoes and peppers returned.
MIT's QCRI solution integrates the pre-cooking phase and determines what makes a delicious and appealing pizza with the right layering. In theory, at least, this could be an AI-based solution for preparing and serving pizza. Pizza isn't the only thing the robot could make, as it understands what the end result of the project should look like.
Although the model is assessed only in the context of pizza, he believes that a similar approach could be useful for foods that are naturally layered, such as meat, vegetables, and even fruits and vegetables. When chilli peppers are cut individually, for example, the ingredients are mesmerizing, from the simple outside to the sauce spewing out of the nozzle.
Recently, I visited the start-up PicnicNews in Seattle to see how its pizza robot can assemble 300 cakes an hour. But most restaurants would be able to make the dough, bake the pizza and serve it in a few minutes, rather than in hours or even days or weeks.
Domino's Pizza, for example, is currently testing a computer vision solution for quality control. At each location, the company uses artificial intelligence to monitor the pizza that comes out of the oven to see if it looks good enough to meet its standards.
The spike is measured and quantified by machine learning to ensure customers don't get a shitty pie. The plate must be measured, evaluated and measured again and again before it is even cooked, so that it is not in the wrong hands and the customer does not have to get it.
The solution from MIT QCRI integrates the pre-cooking step and determines the right amount of cake to prepare a delicious pizza and flatter the eye.
Of course, pizza isn't the only thing a robot can do to understand what the ingredients, steps, and end result should look like, but the idea of an AI robot system that can make pizza on its own is too cool, at least that's the theory. We still have many years to dig a little deeper into this story, because we have had problems serving pizza to our guests for years. The story is a little too delicious for my liking, and I'm still not sure whether the AI robotic system's ability to make its own pizzas was too "cool" or not.
Picnic's modular robotic system will soon transform from a simple pizza preparation to the preparation of human cooks, turning the system into a full-fledged restaurant with its own kitchen and kitchen equipment.
People cannot be 100 per cent consistent, but they will come incredibly close, and Domino's showed last year that it is willing to remove people from the pizza delivery equation. When you think of cutting-edge technology in the world of food distribution and processing, pizza probably comes to mind first.
The squat, pod-shaped robots use cameras and sensors to carry pizzas to hungry customers. This allowed Domino's to collect data on the most popular toppings, including the number, shape and size of each toppings, as well as the size and shape of the pizza itself, and provided the Data Science team with the opportunity to improve its ability to develop an image classification model to identify the toppings and identify the different toppings.
The huge inventory of customer photos offered the Data Science team a unique opportunity for deep learning and machine learning. The development of this in-depth knowledge and expertise has enabled Domino's to apply more advanced classification and prediction models to a wide range of pizza toppings as well as to the pizza itself.