US 12,450,865 B2
Method and system for detecting liquid level inside a container
Agathe Camille Foussat, Singapore (SG); Jiahang Song, Singapore (SG); Enzo Acerbi, Singapore (SG); Laurens Alexander Drapers, Utrecht (NL); and Ruben Zadok Hekster, Utrecht (NL)
Assigned to N.V. Nutricia, Zoetermeer (NL)
Appl. No. 17/783,666
Filed by N.V. Nutricia, Zoetermeer (NL)
PCT Filed Dec. 10, 2019, PCT No. PCT/EP2019/084415
§ 371(c)(1), (2) Date Jun. 9, 2022,
PCT Pub. No. WO2021/115569, PCT Pub. Date Jun. 17, 2021.
Prior Publication US 2022/0375191 A1, Nov. 24, 2022
Int. Cl. G06V 10/44 (2022.01); G06T 7/00 (2017.01); G06V 10/82 (2022.01)
CPC G06V 10/454 (2022.01) [G06T 7/0004 (2013.01); G06V 10/82 (2022.01); G06T 2207/20081 (2013.01); G06T 2207/20084 (2013.01); G06T 2207/30128 (2013.01)] 26 Claims
OG exemplary drawing
 
1. A computer-implemented method of detecting a liquid level inside a container, the method comprising the steps of:
capturing, by a camera of a portable device, a first image of the container;
providing the first image to an input layer of a convolutional neural network, CNN;
obtaining, from a final layer of the CNN, the liquid level inside the container in the first image; and
storing the obtained liquid level, wherein the CNN is configured to identify features of a plurality of volume indicators of the container in the first image and to determine the liquid level in the container in the first image based on the identified features, and
wherein the plurality of volume indicators comprises at least one number and a plurality of scale markings,
wherein the CNN is a modified residual neural network (ResNet) modified by one or more of:
removing a final layer of the ResNet,
inserting classes for classification of liquid level,
inserting additional fully connected layers,
performing dropout in the fully connected layers,
introducing a regularization term,
adding two dimensional convolutions,
applying weight quantization, and
folding convolutional layers onto each other.