Image-based inventory system

1 . A system for image-based inventory determination, the system comprising:

a cart comprising:

a camera; and

a cart computing system in communication with the camera, the cart computing system comprising:

a cart processor; and

a cart memory storing instructions that, when executed by the cart processor, cause the cart computing system to:

capture and store a set of images from the camera;

validate the set of images by:

 inputting each image of the set of images into a machine learning model trained to determine a quality of each image;

 comparing, for each image of the set of images, a respective quality of the image to a threshold quality metric; and

 in response to the respective quality violating the threshold quality metric, deleting the respective image from the set of images;

process the set of images to identify an image of the set of images that displays a section of shelving having both a left vertical beam and a right vertical beam and no additional vertical beams beyond the left vertical beam and the right vertical beam;

locate a label within the identified image, the label identifying a location of the section of shelving;

modify metadata of the identified image to include the location of the section of shelving; and

transmit the identified image to a remote server.

2 . The system of claim 1 , wherein:

the camera comprises a first camera and a second camera, and

the set of images comprises a set of first images from the first camera and a set of second images from the second camera.

3 . The system of claim 2 , wherein the cart further comprises a pole configured to position the first camera at a first height and the second camera at a second height, the first height being greater than the second height.

4 . The system of claim 2 , wherein the identified image is from the set of first images, and further comprises a timestamp, and wherein the memory stores further instructions that, when executed by the processor, cause the computing system to:

identify an associated image from the set of second images sharing the timestamp; and

transmit both the identified image and the associated image to the remote server.

5 . The system of claim 1 , wherein the cart computing system further comprises a display in communication with the cart computing system, the cart computing system providing a graphical user interface (GUI) on the display configured to receive a user input and to issue a command to the camera in response to the user input.

6 . The system of claim 1 , wherein the quality comprises a focus of each of the set of images, and wherein comparing the respective quality of an image to the threshold quality metric comprises evaluating a presence of defined edges in the image.

7 . The system of claim 1 , wherein the machine learning model comprises a blind image quality assessment model that evaluates the quality of each of the set of images without a standard reference image.

8 . The system of claim 1 , wherein:

the camera comprises a video camera;

the set of images comprises a video clip; and

the identified image comprises a single frame of the video clip.

9 . A computer-implemented method for image-based inventory determination, the method comprising:

receiving, by a computing system, an image of a set of images from an optical imaging device, the set of images displaying a series of shelves, the image including a quality metric satisfying a threshold quality metric, wherein the quality metric is determined by a blind image quality assessment model based on the image without using a standard reference image;

determining, by the computing system, that the image of the set of images displays a first vertical beam;

in response to determining that the image displays the first vertical beam, determining, by the computing system, that the image displays a second vertical beam;

in response to determining that the image displays both the first vertical beam and the second vertical beam and that the image does not display any additional vertical beams, identifying, by the computing system, the image as a target image displaying a single shelf;

locating, by the computing system, a label within the target image, the label identifying a location of the single shelf;

modifying, by the computing system, metadata of the target image to include the location of the single shelf; and

transmitting, by the computing system, the target image to a backend computing system.

10 . The method of claim 9 , wherein:

the optical imaging device comprises a video camera;

the set of images comprises a video clip; and

the target image comprises a single frame of the video clip.

11 . The method of claim 9 , wherein:

the optical imaging device comprises a first optical imaging device and a second optical imaging device, and

the set of images comprises a set of first images from the first optical imaging device and a set of second images from the second optical imaging device.

12 . The method of claim 11 , wherein the target image is from the set of first images and further comprises a timestamp, and

wherein the method further comprises:

identifying a second image from the set of second images sharing the timestamp;

associating the second image with the target image; and

transmitting both the target image and the second image to the backend computing system.

13 . The method of claim 12 , further comprising:

determining one or more visual elements shared by the target image and the second image;

combining the target image and the second image by overlapping the shared one or more visual elements; and

transmitting the combined image to the backend computing system.

14 . The method of claim 9 , wherein the target image comprises a first target image,

the single shelf comprises a first single shelf, and the label comprises a first label, and

wherein the method further comprises:

identifying a second target image from the set of images, the second target image displaying a second single shelf;

locating a second label within the second target image, the second label identifying a location of the second single shelf;

comparing the location of the first single shelf to the location of the second single shelf; and

in response to the location of the second single shelf being different than the location of the first single shelf, transmitting the second target image to the backend computing system.

15 . The method of claim 9 , wherein the target image comprises a first target image, the single shelf comprises a first single shelf, and the first target image comprises a first timestamp, and

wherein the method further comprises:

identifying a second target image having a second timestamp from the set of images, the second target image displaying a second single shelf;

determining an amount of time between the first timestamp and the second timestamp; and

in response to the amount of time exceeding a threshold value, transmitting the second target image to the backend computing system.

16 . The method of claim 9 , wherein the optical imaging device is mounted on a cart.

17 . A system for image-based inventory determination, the system comprising:

a cart comprising an optical imaging device; and

a computing system comprising:

a processor; and

a memory storing instructions that, when executed by the processor, cause the computing system to:

receive a video feed from the optical imaging device, the video feed displaying a series of shelving;

identify a target frame from the video feed that displays two vertical shelf beams and does not display any additional vertical shelf beams, wherein the target frame includes a quality metric satisfying a threshold quality metric;

locate a label within the target frame, the label identifying a location of the series of shelving;

modify metadata of the target frame to include the location of the series of shelving; and

transmit the target frame to a remote server.

18 . The system of claim 17 , wherein identifying the target frame comprises:

determining that a single frame of the video feed displays a first vertical beam;

in response to determining that the single frame displays the first vertical beam, determining that the single frame displays a second vertical beam; and

in response to determining that the image displays both the first vertical beam and the second vertical beam, identifying the single frame as the target frame displaying a single shelf.

19 . The system of claim 17 , wherein the memory further stores instructions that, when executed by the processor, cause the computing system to:

locate a label within the target frame, the label identifying a location of a single shelf defined by the two vertical shelf beams; and

associating the location with the target image.

20 . The system of claim 19 , wherein the target frame comprises a first target frame, the single shelf comprises a first single shelf, and the label comprises a first label, and

wherein the memory further stores instructions that, when executed by the processor, cause the computing system to:

identify a second target frame from the video feed, the second target frame displaying a second single shelf;

locating a second label within the second target frame, the second label identifying a location of the second single shelf;

comparing the location of the first single shelf to the location of the second single shelf; and

in response to the location of the second single shelf being different than the location of the first single shelf, transmitting the second target frame to the remote server.