Predictive classification of insects
Insects can be classified into a category (e.g., sex category, species category, size category, etc.) using a variety of different classification approaches including, for example, an industrial vision classifier and/or a machine learning classifier. At least some classification approaches may be used in real-time to make real-time decisions and others can be used to validate earlier-made real-time decisions.
1. A system, comprising:
an imaging device configured to capture images of insects; and
a computing device in communication with the imaging device, and configured to at least:
instruct the imaging device to capture an image that depicts at least a portion of an insect of a plurality of insects at a first location, wherein each of the plurality of insects is moved to the first location using a movement device configured to singulate the plurality of insects;
receive, from the imaging device, the image;
determine, using an industrial vision classifier:
(i) a first classification of the image into at least one category based at least in part on features extracted from the image; and
(ii) a first confidence measure corresponding to the first classification;
execute an instruction to cause movement of the insect to a second location based at least in part on the first classification and the first confidence measure;
after causing movement of the insect to the second location, determine, using a machine learning classifier:
(i) a second classification of the image into the at least one category based at least in part on the image; and
(ii) a second confidence measure corresponding to the second classification; and
determine a third classification of the image based at least in part on the first confidence measure and the second confidence measure.
2. The system of claim 1 , wherein each of the first classification, the second classification, and the third classification comprises a classification of the image into the at least one category that is based on insect sex.
3. The system of claim 1 , wherein the computing device is further configured to at least execute an instruction to cause movement of the insect from the second location to a third location based at least in part on the third classification.
4. The system of claim 3 , wherein the first location is an insect sortation device, the second location is an insect transportation device, and the third location is outside of the insect transportation device.
5. The system of claim 1 , wherein:
the insect is located at the second location while the third classification of the image is determined; and
the computing device is further configured to at least execute an instruction to cause movement of the insect to a third location based at least in part on the third classification.
6. The system of claim 5 , wherein the second location is an insect sortation device and the third location is an insect transportation device.
7. The system of claim 1 , wherein:
the image comprises a sequence of images and the first confidence measure comprises a plurality of first confidence measures corresponding to individual images of the sequence of images; and
the second confidence measure comprises a plurality of second confidence measures corresponding to the individual images of the sequence of images.
8. The system of claim 1 , wherein the third classification of the image corresponds to one category of a plurality of categories comprising male, female, intersex, species, species, or gynandromorph.
9. The system of claim 1 , wherein the insect comprises an adult mosquito.
10. The system of claim 1 , wherein determining the third classification comprises determining a composite classification based on the first classification and the second classification.
11. The system of claim 10 , wherein the composite classification comprises a composite confidence measure.
12. The system of claim 1 , wherein the machine learning classifier comprises a core deep neural network and a recurrent neural network model.
13. The system of claim 12 , wherein the core deep neural network is a trained core deep neural network, the training comprising:
accessing a set of labeled training images, each individual training image including an image of a singulated insect and a label, wherein the label is associated with a label frequency;
determining, using a loss function, a penalty proportional to the inverse of the label frequency; and
updating at least one weight of the core deep neural network using the penalty and based on a stochastic gradient descent and backpropagation technique.
14. A computer-implemented method, comprising:
instructing, an imaging device, to capture an image that depicts at least a portion of an insect of a plurality of insects at a first location, wherein each of the plurality of insects is moved to the first location using a movement device configured to singulate the plurality of insects;
receiving, from the imaging device, the image;
determining, using an industrial vision classifier:
(i) a first classification of the image into at least one category based at least in part on features extracted from the image; and
(ii) a first confidence measure corresponding to the first classification;
executing an instruction to cause movement of the insect to a second location based at least in part on the first classification and the first confidence measure;
after causing movement of the insect to the second location, determining, using a machine learning classifier:
(i) a second classification of the image into the at least one category based at least in part on the image; and
(ii) a second confidence measure corresponding to the second classification; and
determining a third classification of the image based at least in part on the first confidence measure and the second confidence measure.
15. The computer-implemented method of claim 14 , wherein each of the first classification, the second classification, and the third classification comprises a classification of the image into the at least one category that is based on insect sex.
16. The computer-implemented method of claim 14 , further comprising executing an instruction to cause movement of the insect from the second location to a third location based at least in part on the third classification.
17. The computer-implemented method of claim 16 , wherein the first location is an insect sortation device, the second location is an insect transportation device, and the third location is outside of the insect transportation device.
18. The computer-implemented method of claim 14 , wherein:
the insect is located at the second location while the third classification of the image is determined; and
the method further comprises executing an instruction to cause movement of the insect to a third location based at least in part on the third classification.
19. The computer-implemented method of claim 18 , wherein the second location is an insect sortation device and the third location is an insect transportation device.
20. The computer-implemented method of claim 14 , wherein:
the image comprises a sequence of images and the first confidence measure comprises a plurality of first confidence measures corresponding to individual images of the sequence of images; and
the second confidence measure comprises a plurality of second confidence measures corresponding to the individual images of the sequence of images.
21. The computer-implemented method of claim 14 , wherein the third classification of the image corresponds to one category of a plurality of categories comprising male, female, intersex, species, species, or gynandromorph.
22. The computer-implemented method of claim 14 , wherein the insect comprises an adult mosquito.
23. The computer-implemented method of claim 14 , wherein machine learning classifier comprises a core deep neural network and a recurrent neural network model.
24. The computer-implemented method of claim 23 , wherein the core deep neural network is a trained core deep neural network, the training comprising:
accessing a set of labeled training images, each individual training image including an image of a singulated insect and a label, wherein the label is associated with a label frequency;
determining, using a loss function, a penalty proportional to the inverse of the label frequency; and
updating at least one weight of the core deep neural network using the penalty and based on a stochastic gradient descent and backpropagation technique.
25. One or more non-transitory computer-readable media comprising computer-executable instructions that, when executed by one or more computing systems, cause the one or more computing systems to:
instruct, an imaging device, to capture an image that depicts at least a portion of an insect of a plurality of insects at a first location, wherein each of the plurality of insects is moved to the first location using a movement device configured to singulate the plurality of insects;
receive, from the imaging device, the image;
determine, using an industrial vision classifier:
(i) a first classification of the image into at least one category based at least in part on features extracted from the image; and
(ii) a first confidence measure corresponding to the first classification;
execute an instruction to cause movement of the insect to a second location based at least in part on the first classification and the first confidence measure;
after causing movement of the insect to the second location, determine, using a machine learning classifier:
(i) a second classification of the image into the at least one category based at least in part on the image; and
(ii) a second confidence measure corresponding to the second classification; and
determine a third classification of the image based at least in part on the first confidence measure and the second confidence measure.
26. The one or more non-transitory computer-readable media of claim 25 , wherein each of the first classification, the second classification, and the third classification comprises a classification of the image into the at least one category that is based on insect sex.
27. The one or more non-transitory computer-readable media of claim 25 , further comprising additional computer-executable instructions that, when executed by the one or more computing systems, cause the one or more computing systems to execute an instruction to cause movement of the insect from the second location to a third location based at least in part on the third classification.
28. The one or more non-transitory computer-readable media of claim 27 , wherein the first location is an insect sortation device, the second location is an insect transportation device, and the third location is outside of the insect transportation device.
29. The one or more non-transitory computer-readable media of claim 25 , wherein the insect is located at the second location while the third classification of the image is determined; and further comprising additional computer-executable instructions that, when executed by the one or more computing systems, cause the one or more computing systems to execute an instruction to cause movement of the insect to a third location based at least in part on the third classification.
30. The one or more non-transitory computer-readable media of claim 29 , wherein the second location is an insect sortation device and the third location is an insect transportation device.
31. The one or more non-transitory computer-readable media of claim 25 , wherein:
the image comprises a sequence of images and the first confidence measure comprises a plurality of first confidence measures corresponding to individual images of the sequence of images; and
the second confidence measure comprises a plurality of second confidence measures corresponding to the individual images of the sequence of images.
32. The one or more non-transitory computer-readable media of claim 25 , wherein the third classification of the image corresponds to one category of a plurality of categories comprising male, female, intersex, species, species, or gynandromorph.
33. The one or more non-transitory computer-readable media of claim 25 , wherein the insect comprises an adult mosquito.
34. The one or more non-transitory computer-readable media of claim 25 , wherein the machine learning classifier comprises a core deep neural network and a recurrent neural network model.
35. The one or more non-transitory computer-readable media of claim 34 , wherein the core deep neural network is a trained core deep neural network, the training comprising:
accessing a set of labeled training images, each individual training image including an image of a singulated insect and a label, wherein the label is associated with a label frequency;
determining, using a loss function, a penalty proportional to the inverse of the label frequency; and
updating at least one weight of the core deep neural network using the penalty and based on a stochastic gradient descent and backpropagation technique.