Robot chef can learn recipes from YouTube


Researchers at Cambridge University have trained a robot chef that learns how to recreate recipes by watching cooking videos.

The robot uses a machine learning method known as “imitation learning” to understand the actions in the video and then tries to replicate those actions to create the same dish.

Researchers programmed the robot with a “cookbook” of eight simple salad recipes. After watching a video of a human demonstrating one of the recipes, the robo-chef could identify the dish being prepared and recreate it.

ADVERTISEMENT

By the end of the experiment, it was even able to create its own recipe and add it to the pre-programmed cookbook, Cambridge University said in a press release.

Of the 16 videos it watched, the robot correctly identified the recipe 93% of the time, even though it only detected 83% of the human chef’s actions, researchers said.

The robot could also understand that slight variations in a recipe, such as making a double portion of a dish, were just variations and not new recipes.

“It’s amazing how much nuance the robot was able to detect,” Grzegorz Sochacki, the paper’s first author, said.

But the robot chef also has its limits. For example, it struggled to identify a carrot if the human demonstrator wrapped their hand around it.

Recreating viral food videos posted online by social media influencers would also be difficult because of the fast cuts and visual effects used in such clips.

“Our robot isn’t interested in the sorts of food videos that go viral on social media – they’re simply too hard to follow,” said Sochacki.

ADVERTISEMENT

“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,” he added.

The results, published in the IEEE Access journal, highlight the potential of video content as a rich data source for automated food production, potentially enabling easier and cheaper deployment of robot chefs, researchers said.

The training process of a robot chef included using a publicly-available neural network and computer vision techniques, according to Cambridge University.

The robot analyzed each frame of the video, identifying different objects and features such as the ingredients, a knife, and a human demonstrator’s arms, hands, and face, it said.