MSGID: 1:317/3 647eb6a2   
   PID: hpt/lnx 1.9.0-cur 2019-01-08   
   TID: hpt/lnx 1.9.0-cur 2019-01-08   
    Robot 'chef' learns to recreate recipes from watching food videos    
      
     Date:   
         June 5, 2023   
     Source:   
         University of Cambridge   
     Summary:   
         Researchers have trained a robotic 'chef' to watch and learn from   
         cooking videos, and recreate the dish itself.   
      
      
         Facebook Twitter Pinterest LinkedIN Email   
      
   ==========================================================================   
   FULL STORY   
   ==========================================================================   
   Researchers have trained a robotic 'chef' to watch and learn from cooking   
   videos, and recreate the dish itself.   
      
   The researchers, from the University of Cambridge, programmed their   
   robotic chef with a 'cookbook' of eight simple salad recipes. After   
   watching a video of a human demonstrating one of the recipes, the robot   
   was able to identify which recipe was being prepared and make it.   
      
   In addition, the videos helped the robot incrementally add to its   
   cookbook. At the end of the experiment, the robot came up with a ninth   
   recipe on its own.   
      
   Their results, reported in the journal IEEE Access, demonstrate how video   
   content can be a valuable and rich source of data for automated food   
   production, and could enable easier and cheaper deployment of robot chefs.   
      
   Robotic chefs have been featured in science fiction for decades, but in   
   reality, cooking is a challenging problem for a robot. Several commercial   
   companies have built prototype robot chefs, although none of these are   
   currently commercially available, and they lag well behind their human   
   counterparts in terms of skill.   
      
   Human cooks can learn new recipes through observation, whether   
   that's watching another person cook or watching a video on YouTube,   
   but programming a robot to make a range of dishes is costly and   
   time-consuming.   
      
   "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, the paper's first author.   
      
   Sochacki, a PhD candidate in Professor Fumiya Iida's Bio-Inspired   
   Robotics Laboratory, and his colleagues devised eight simple salad   
   recipes and filmed themselves making them. They then used a publicly   
   available neural network to train their robot chef. The neural network   
   had already been programmed to identify a range of different objects,   
   including the fruits and vegetables used in the eight salad recipes   
   (broccoli, carrot, apple, banana and orange).   
      
   Using computer vision techniques, the robot analysed each frame of video   
   and was able to identify the different objects and features, such as a   
   knife and the ingredients, as well as the human demonstrator's arms, hands   
   and face. Both the recipes and the videos were converted to vectors and   
   the robot performed mathematical operations on the vectors to determine   
   the similarity between a demonstration and a vector.   
      
   By correctly identifying the ingredients and the actions of the   
   human chef, the robot could determine which of the recipes was being   
   prepared. The robot could infer that if the human demonstrator was   
   holding a knife in one hand and a carrot in the other, the carrot would   
   then get chopped up.   
      
   Of the 16 videos it watched, the robot recognised the correct recipe   
   93% of the time, even though it only detected 83% of the human chef's   
   actions. The robot was also able to detect that slight variations in   
   a recipe, such as making a double portion or normal human error, were   
   variations and not a new recipe. The robot also correctly recognised the   
   demonstration of a new, ninth salad, added it to its cookbook and made it.   
      
   "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 recognising, for example,   
   that two chopped apples and two chopped carrots is the same recipe as   
   three chopped apples and three chopped carrots."  The videos used to   
   train the robot chef are not like the food videos made by some social   
   media influencers, which are full of fast cuts and visual effects, and   
   quickly move back and forth between the person preparing the food and the   
   dish they're preparing. For example, the robot would struggle to identify   
   a carrot if the human demonstrator had their hand wrapped around it --   
   for the robot to identify the carrot, the human demonstrator had to hold   
   up the carrot so that the robot could see the whole vegetable.   
      
   "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. "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."  The research was supported in part by Beko   
   plc and the Engineering and Physical Sciences Research Council (EPSRC),   
   part of UK Research and Innovation (UKRI).   
      
       * RELATED_TOPICS   
             o Plants_&_Animals   
                   # Food # Food_and_Agriculture # Agriculture_and_Food   
             o Matter_&_Energy   
                   # Robotics_Research # Engineering # Vehicles   
             o Computers_&_Math   
                   # Robotics # Artificial_Intelligence # Neural_Interfaces   
       * RELATED_TERMS   
             o Robotic_surgery o Dog_intelligence o Cat_intelligence o   
             Fullerene o Potato o Cherry o Scallop o Swordfish   
      
   ==========================================================================   
   Story Source: Materials provided by University_of_Cambridge. The original   
   text of this story is licensed under a Creative_Commons_License. Note:   
   Content may be edited for style and length.   
      
      
   ==========================================================================   
   Journal Reference:   
      1. Grzegorz Sochacki, Arsen Abdulali, Narges Khadem Hosseini,   
      Fumiya Iida.   
      
         Recognition of Human Chef's Intentions for Incremental Learning of   
         Cookbook by Robotic Salad Chef. IEEE Access, 2023; 1 DOI: 10.1109/   
         ACCESS.2023.3276234   
   ==========================================================================   
      
   Link to news story:   
   https://www.sciencedaily.com/releases/2023/06/230605181344.htm   
      
   --- up 1 year, 14 weeks, 10 hours, 51 minutes   
    * Origin: -=> Castle Rock BBS <=- Now Husky HPT Powered! (1:317/3)   
   SEEN-BY: 15/0 106/201 114/705 123/120 153/7715 218/700 226/30 227/114   
   SEEN-BY: 229/110 112 113 307 317 400 426 428 470 664 700 291/111 292/854   
   SEEN-BY: 298/25 305/3 317/3 320/219 396/45   
   PATH: 317/3 229/426