US 12,277,531 B2
Method and system for classifying food items
Marc Zornes, London (GB); Kevin Duffy, London (GB); David Woosnam, London (GB); Peter Leonard Krebs, Iowa City, IA (US); Minh-Tri Pham, London (GB); Phong Vo, Birmingham (GB); and Mark Haynes, Wandsworth (GB)
Assigned to Winnow Solutions Limited, Milton Keynes (GB)
Appl. No. 16/967,966
Filed by WINNOW SOLUTIONS LIMITED, Milton Keynes (GB)
PCT Filed Feb. 7, 2019, PCT No. PCT/GB2019/050338
§ 371(c)(1), (2) Date Aug. 6, 2020,
PCT Pub. No. WO2019/155220, PCT Pub. Date Aug. 15, 2019.
Claims priority of application No. 1802022 (GB), filed on Feb. 7, 2018.
Prior Publication US 2021/0397648 A1, Dec. 23, 2021
Int. Cl. G06Q 10/0875 (2023.01); G06F 16/55 (2019.01); G06F 18/24 (2023.01); G06N 20/00 (2019.01); G06V 30/194 (2022.01)
CPC G06Q 10/0875 (2013.01) [G06F 16/55 (2019.01); G06F 18/24 (2023.01); G06N 20/00 (2019.01); G06V 30/194 (2022.01)] 28 Claims
OG exemplary drawing
 
1. A method for classifying food items, including:
capturing one or more sensor data relating to a disposal event, wherein the disposal event includes a food item being placed into a waste receptacle by a user, and the disposal event is one of a plurality of disposal events relating to the waste receptacle such that food items from multiple events are disposed of consecutively within the waste receptacle without the waste receptacle being emptied, wherein the one or more sensor data includes image data captured from an image sensor above the waste receptacle and wherein the image data includes image data captured while the food item is being placed within the waste receptacle or after the food item has been placed within the waste receptacle onto food items within one or more previous disposal events, and wherein the one or more sensor data includes weight data;
processing the image data to isolate new objects within the image data compared to previously captured image data using a segmenter; and
classifying the food item using at least the processed image data, at least in part, automatically using a model trained on sensor data, and using the weight data.