Automated image curation for machine learning deployments
The present disclosure provides techniques for data curation and image evaluation. A first image is captured, and a first indication of a first item is received. A first identifier of the first item is then identified based on the first indication. Further, based on the first indication, it is determined that the first image depicts the first item. The first image is labeled with the first identifier, and a machine learning (ML) model of an ML system is trained based on the labeled first image.
1. A method, comprising:
capturing a first image;
receiving a first indication of a first item;
identifying, based on the first indication, a first identifier of the first item;
determining, based on the first indication, that the first image depicts the first item;
labeling the first image with the first identifier;
training a machine learning (ML) model of an ML system based on the labeled first image;
capturing a second image;
generating a predicted identifier based on processing the second image using the ML model;
determining that the ML system is in a verification state;
receiving a second indication of the first item;
identifying, based on the second indication, the first identifier of the first item; and
storing the second image, along with the predicted identifier and the first identifier.
2. The method of claim 1 , further comprising:
training the ML model based on the labeled first image based on determining that the ML system is in a training phase.
3. The method of claim 1 , the method further comprising:
determining that the predicted identifier and the first identifier do not match; and
refining the ML model based on the second image and the first identifier.
4. The method of claim 1 , the method further comprising:
determining that the ML system is in an active state;
refraining from receiving a third indication of the first item; and
logging the predicted identifier.
5. The method of claim 1 , wherein determining, based on the first indication, that the first image depicts the first item comprises:
identifying a timestamp associated with the first indication;
identifying a location associated with the first indication; and
determining that the timestamp and location associated with the first indication align with a timestamp and location associated with the first image.
6. The method of claim 1 , wherein determining, based on the first indication, that the first image depicts the first item comprises:
identifying a location associated with the first indication; and
in response to receiving the first indication, causing the first image to be captured of the location associated with the first indication.
7. A computer-readable storage medium containing computer program code that, when executed by operation of one or more computer processors, performs an operation comprising:
capturing a first image;
receiving a first indication of a first item;
identifying, based on the first indication, a first identifier of the first item;
determining, based on the first indication, that the first image depicts the first item;
labeling the first image with the first identifier;
training a machine learning (ML) model of an ML system based on the labeled first image;
capturing a second image;
generating a predicted identifier based on processing the second image using the ML model;
determining that the ML system is in a verification state;
receiving a second indication of the first item;
identifying, based on the second indication, the first identifier of the first item; and
storing the second image, along with the predicted identifier and the first identifier.
8. The computer-readable storage medium of claim 7 , the operation further comprising:
determining that the predicted identifier and the first identifier do not match; and
refining the ML model based on the second image and the first identifier.
9. The computer-readable storage medium of claim 7 , the operation further comprising:
determining that the ML system is in an active state;
refraining from receiving a third indication of the first item; and
logging the predicted identifier.
10. The computer-readable storage medium of claim 7 , wherein determining, based on the first indication, that the first image depicts the first item comprises:
identifying a timestamp associated with the first indication;
identifying a location associated with the first indication; and
determining that the timestamp and location associated with the first indication align with a timestamp and location associated with the first image.
11. A system comprising:
one or more computer processors; and
a memory containing a program which when executed by the one or more computer processors performs an operation, the operation comprising:
capturing a first image;
receiving a first indication of a first item;
identifying, based on the first indication, a first identifier of the first item;
determining, based on the first indication, that the first image depicts the first item;
labeling the first image with the first identifier;
training a machine learning (ML) model of an ML system based on the labeled first image;
capturing a second image;
generating a predicted identifier based on processing the second image using the ML model;
determining that the ML system is in a verification state;
receiving a second indication of the first item;
identifying, based on the second indication, the first identifier of the first item; and
storing the second image, along with the predicted identifier and the first identifier.
12. The system of claim 11 , the operation further comprising:
determining that the predicted identifier and the first identifier do not match; and
refining the ML model based on the second image and the first identifier.
13. The system of claim 11 , the operation further comprising:
determining that the ML system is in an active state;
refraining from receiving a third indication of the first item; and
logging the predicted identifier.
14. The system of claim 11 , wherein determining, based on the first indication, that the first image depicts the first item comprises:
identifying a timestamp associated with the first indication;
identifying a location associated with the first indication; and
determining that the timestamp and location associated with the first indication align with a timestamp and location associated with the first image.