Robots are becoming increasingly important in different industries and are being trained worldwide to perform various tasks with great precision. including cleaning and even playing football.
Although robots still have a lot to learn when it comes to cooking, they are quickly catching up. Researchers from the University of Cambridge have taken a significant step forward in this area. They have trained a robot chef to recreate recipes simply by watching cooking videos.
In this experiment, the robot was taught to replicate eight simple salad recipes. The researchers filmed themselves making these salads to provide training material for the robot. Surprisingly, the videos also helped the robot create its own recipe. By using all the available resources, the robot developed a ninth recipe on its own.
“We wanted to see whether we could train a robot chef to learn in the same incremental way that humans can – by identifying the ingredients and how they go together in the dish,” said Grzegorz Sochacki from Cambridge’s Department of Engineering, and the paper’s first author, in an official press statement.
To program the robotic chef, the team utilized publicly available neural network algorithms. This allowed the robot to learn how to recognize different objects, such as fruits, vegetables, and even the human demonstrator’s knife, arms, hands, and faces.
By watching a total of 16 videos, the robot successfully identified the recipe with an accuracy of 93 percent. This experiment demonstrates that video content can be a valuable resource for training robots to automate food production. In the future, robots might assist human chefs in preparing food in various hospitality settings.
“It’s amazing how much nuance the robot was able to detect,” said Sochacki. “These recipes aren’t complex – they’re essentially chopped fruits and vegetables, but it was really effective at recognizing, for example, that two chopped apples and two chopped carrots are the same recipe as three chopped apples and three chopped carrots.”
The results have been published in the journal IEEE Access.
