Detecting cooking state with gas sensors during dry cooking

Gas sensors have the potential to assist cooking by providing feedback on the cooking process and by further automating cooking. In this work, we explored the potential use of gas sensors to monitor food during the cooking process. Focusing on dry cooking, we collected gas emissions using 13 sensors during trials in which food was cooked to various degrees of doneness. Using decision tree classifiers, we were able to predict doneness for waffles and popcorn with 73% and 85% accuracy, respectively. We reflect on the potential reasons for this variation and the ways in which gas sensors might reliably be used in ubicomp applications to support cooking.