Automatic labeling of products via expedited checkout system
A portable checkout unit automatically generates training data for an automatic checkout system as a customer collects items in a store. A customer uses an item scanner of portable checkout unit to generate a virtual shopping list of items collected in the shopping cart. When the customer adds a new item to the shopping cart or on some regular interval, the portable checkout unit captures images of the items contained by the shopping cart and can generate bounding boxes for each product in each image. The bounding boxes can be associated with item identifiers from previously-generated bounding boxes to identify the items captured by the bounding boxes. Each bounding box paired with an item identifier can then be used as training data for an automated checkout system.
1. A method comprising:
detecting, at a first time, that a first item was added to a first shopping cart;
capturing a first image of the first shopping cart comprising the first item;
applying the first image to a machine-learned model, wherein the machine learned model was trained to output an item identifier for the first item by:
responsive to detecting that a second item was added to a second shopping cart at a second time prior to the first time:
capturing a second image using a camera attached to the second shopping cart, the camera being directed towards a plurality of items contained by the shopping cart including the second item, the second item having a same item identifier as the first item;
identifying, for each item of the plurality of items, one or more portions of the second image that contain the item; and
generating training data that comprises, for each given item of the plurality of items, the identified one or more portions of the second image that contain the given item, and a label comprising an item identifier for the given item;
receiving, as output from the machine-learned model, the item identifier for the first item; and
updating a virtual shopping list based on the received item identifier.
2. The method of claim 1 , wherein an item identifier for the second item was determined using an item scanner of a portable checkout unit when the second item was added to the second shopping cart.
3. The method of claim 1 , further comprising: receiving sensor data from one or more sensors attached to the shopping cart.
4. The method of claim 3 , wherein the one or more sensors comprise at least one of a pressure sensor, a proximity sensor, a motion sensor, a depth sensor, or an RFID sensor.
5. The method of claim 3 , wherein the generated training data further comprises the received sensor data.
6. The method of claim 3 , further comprising: detecting that the second item was added to the second shopping cart based on the received sensor data.
7. The method of claim 1 , wherein each of the one or more portions comprises a bounding box.
8. The method of claim 7 , wherein each bounding box for the one or more portions is generated based on previously-generated bounding boxes.
9. The method of claim 8 , wherein each bounding box for the one or more portions is paired with a previously-generated bounding box, and wherein each bounding box for the one or more portions is associated with an item identifier associated with the previously-generated bounding box associated with the bounding box.
10. A non-transitory computer-readable medium comprising memory with instructions encoded thereon, the instructions, when executed by one or more processors, causing the one or more processors to perform operations, the instructions comprising instructions to:
detect, at a first time, that a first item was added to a first shopping cart;
capture a first image of the first shopping cart comprising the first item;
apply the first image to a machine-learned model, wherein the machine learned model was trained to output an item identifier for the first item by:
responsive to detecting that a second item was added to a second shopping cart at a second time prior to the first time:
capturing a second image using a camera attached to the second shopping cart, the camera being directed towards a plurality of items contained by the shopping cart including the second item, the second item having a same item identifier as the first item;
identifying, for each item of the plurality of items, one or more portions of the second image that contain the item; and
generating training data that comprises, for each given item of the plurality of items, the identified one or more portions of the second image that contain the given item, and a label comprising an item identifier for the given item;
receive, as output from the machine-learned model, the item identifier for the first item; and
update a virtual shopping list based on the received item identifier.
11. The non-transitory computer-readable medium of claim 10 , wherein an item identifier for the second item was determined using an item scanner of a portable checkout unit when the second item was added to the second shopping cart.
12. The non-transitory computer-readable medium of claim 10 , the instructions further comprising instructions to receive sensor data from one or more sensors attached to the shopping cart.
13. The non-transitory computer-readable medium of claim 12 , wherein the one or more sensors comprise at least one of a pressure sensor, a proximity sensor, a motion sensor, a depth sensor, or an RFID sensor.
14. The non-transitory computer-readable medium of claim 12 , wherein the generated training data further comprises the received sensor data.
15. The non-transitory computer-readable medium of claim 12 , the further comprising instructions to: detect that the second item was added to the second shopping cart based on the received sensor data.
16. The non-transitory computer-readable medium of claim 10 , wherein each of the one or more portions comprises a bounding box.
17. The non-transitory computer-readable medium of claim 16 , wherein each bounding box for the one or more portions is generated based on previously-generated bounding boxes.
18. The non-transitory computer-readable medium of claim 17 , wherein each bounding box for the one or more portions is paired with a previously-generated bounding box, and wherein each bounding box for the one or more portions is associated with an item identifier associated with the previously-generated bounding box associated with the bounding box.
19. A system comprising:
memory with instructions encoded thereon; and
one or more processors that, when executing the instructions, are caused to perform operations comprising:
detecting, at a first time, that a first item was added to a first shopping cart;
capturing a first image of the first shopping cart comprising the first item;
applying the first image to a machine-learned model, wherein the machine learned model was trained to output an item identifier for the first item by:
responsive to detecting that a second item was added to a second shopping cart at a second time prior to the first time:
capturing a second image using a camera attached to the second shopping cart, the camera being directed towards a plurality of items contained by the shopping cart including the second item, the second item having a same item identifier as the first item;
identifying, for each item of the plurality of items, one or more portions of the second image that contain the item; and
generating training data that comprises, for each given item of the plurality of items, the identified one or more portions of the second image that contain the given item, and a label comprising an item identifier for the given item;
receiving, as output from the machine-learned model, the item identifier for the first item; and
updating a virtual shopping list based on the received item identifier.
20. The system of claim 19 , wherein an item identifier for the second item was determined using an item scanner of a portable checkout unit when the second item was added to the second shopping cart.