Engineers from the University of Cambridge have developed a robot chef capable of making salads after watching cooking videos.
Cooking is a complex task for robots, and some humans.
The experiment demonstrates the potential for robots to recreate new recipes after viewing cooking videos, in the same way a person might learn a new dish from Instagram or YouTube.
Lead author Grzegorz Sochacki from Cambridge University Bio-inspired Robotics Laboratory says, “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.”
Sochacki and colleagues devised eight simple salad recipes containing a maximum of five ingredients (broccoli, carrot, apple, banana and orange) and filmed themselves making each salad.
They then used a publicly available artificial intelligence algorithm to train their robot chef using the videos.
After analysing each video – frame by frame – using computer vision, correctly identifying the ingredients and the actions of the human chef, the robot was able to determine which of the recipes was being prepared.
Read more about robots making recipes: 3D printed cheesecake – food printing technology puts the cherry on top
“Firstly, the robot watches the demonstration of a human making a specific salad and compares its similarity to all recipes in the cookbook […] the robotic chef proceeds to prepare a salad according to the recipe with the highest similarity,” says the paper outlining the experiment, published in IEEE Access.
When a new salad was demonstrated, the robot identifies the lack of similarity with recipes already in its cookbook, and adds a new recipe. “This new recipe can now be cooked by the robotic chef with-out further programming,” the authors write.
Of the 16 videos it watched, the robot recognised the correct recipe 93% of the time, detecting 83% of the human chef’s actions.
The robot chef also correctly recognised when a new, ninth salad was being prepared, adding this to its cookbook.
While cooking videos proliferate on social media platforms, the videos used to train the robot chef differ from popular food videos.
“Our robot isn’t interested in the sorts of food videos that go viral on social media – they’re simply too hard to follow,” says Sochacki. “But as these robot chefs get better and faster at identifying ingredients in food videos, they might be able to use sites like YouTube to learn a whole range of recipes.”
Do you care about the oceans? Are you interested in scientific developments that affect them? Then our new email newsletter Ultramarine, launching soon, is for you. Click here to become an inaugural subscriber.