Computationally efficient quality assurance inspection processes using machine learning
Data is received that includes a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system. A representation is generated for each image using a first machine learning model. One or more second machine learning models are then used to analyze each image using the corresponding representation. The analyses can be provided to a consuming application or process for quality assurance analysis.
1. A computer-implemented method for providing quality assurance comprising:
receiving data comprising a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system;
generating, using a first machine learning model, a representation for each image;
analyzing, by one or more second machine learning models, each image using the corresponding representation, the one or more second machine learning models comprise an ensemble of different image analysis inspection tools each trained to characterize a different aspect of the objects; and
providing the analyses to a consuming application or process for quality assurance analysis.
2. The method of claim 1 , wherein the representation is an embedding.
3. The method of claim 1 , wherein the representation comprises one or more labels.
4. The method of claim 1 , wherein the representation comprises one or more attributes.
5. The method of claim 1 , wherein the first machine learning model comprises a vision transformer.
6. The method of claim 5 , wherein the vision transformer is trained using self-supervised learning that provides self-distillation with no labels (DINO).
7. The method of claim 1 , wherein at least a portion of the one or more second machine learning models comprise neural networks.
8. The method of claim 7 , wherein the neural networks have five or fewer layers.
9. The method of claim 7 , wherein the neural networks have three layers.
10. The method of claim 7 , wherein at least one of the neural networks comprises a multilayer perceptron.
11. The method of claim 1 , wherein the representation is constrained within a region of interest of each image.
12. The method of claim 1 , wherein the one or more second machine learning models further comprise an ensemble of different image analysis inspection tools each trained to characterize a same aspect of the objects.
13. The method of claim 1 , wherein the generation of the representation by the first machine learning model is more computationally expensive than the analysis by the one or more second machine learning models.
14. A system for providing quality assurance comprising:
at least one data processor; and
memory comprising instructions which, when executed by the at least one data processor, result in operations comprising:
receiving data comprising a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system;
generating, using a first machine learning model, a representation for each image;
analyzing, by one or more second machine learning models comprising an ensemble of different image analysis tools each trained to characterize one or more aspects of the objects, each image using the corresponding representation; and
providing the analyses to a consuming application or process for quality assurance analysis.
15. The system of claim 14 , wherein the representation is an embedding.
16. The system of claim 14 , wherein the representation comprises one or more labels.
17. The system of claim 14 , wherein the representation comprises one or more attributes.
18. The system of claim 14 , wherein the first machine learning model comprises a vision transformer.
19. The system of claim 18 , wherein the vision transformer is trained using self-supervised learning that provides self-distillation with no labels (DINO).
20. The system of claim 14 , wherein at least a portion of the one or more second machine learning models comprise neural networks.
21. The system of claim 20 , wherein the neural networks have five or fewer layers.
22. The system of claim 20 , wherein the neural networks have three layers.
23. The system of claim 20 , wherein at least one of the neural networks comprises a multilayer perceptron.
24. The system of claim 14 , wherein the representation is constrained within a region of interest of each image.
25. The system of claim 14 , wherein the ensemble of different image analysis inspection tools are each trained to characterize a different aspect of the objects.
26. The system of claim 14 , wherein the ensemble of different image analysis inspection tools are each trained to characterize a same aspect of the objects.
27. The system of claim 14 , wherein the generation of the representation by the first machine learning model is more computationally expensive than the analysis by the one or more second machine learning models.
28. A computer-implemented method for providing quality assurance comprising:
receiving data comprising a feed of images of a plurality of objects passing in front of an inspection camera module forming part of a quality assurance inspection system;
generating, using a vision transformer trained using self-supervised learning that provides self-distillation with no labels (DINO), a representation for each image;
analyzing, by one or more image analysis inspection tools, each image using the corresponding representation; and
providing the analyses to a consuming application or process for quality assurance analysis.
29. The method of claim 28 , wherein the vision transformer comprises a neural network having only three layers.