Using empirical evidence to generate synthetic training data for plant detection
Implementations are described herein for automatically generating synthetic training images that are usable as training data for training machine learning models to detect, segment, and/or classify various types of plants in digital images. In various implementations, a digital image may be obtained that captures an area. The digital image may depict the area under a lighting condition that existed in the area when a camera captured the digital image. Based at least in part on an agricultural history of the area, a plurality of three-dimensional synthetic plants may be generated. The synthetic training image may then be generated to depict the plurality of three-dimensional synthetic plants in the area. In some implementations, the generating may include graphically incorporating the plurality of three-dimensional synthetic plants with the digital image based on the lighting condition.
1. A method for generating a synthetic training image, the method implemented using one or more processors and comprising:
obtaining a digital image that captures an area;
based at least in part on an agricultural history of the area, generating a plurality of three-dimensional synthetic plants; and
generating the synthetic training image to depict the plurality of three-dimensional synthetic plants in the area, wherein the generating includes graphically incorporating the plurality of three-dimensional synthetic plants with the digital image at a spacing that is determined based at least in part on a feature of the area when the area was captured in the digital image.
2. The method of claim 1 , wherein the feature of the area comprises an environmental condition of the area.
3. The method of claim 2 , wherein the environmental condition of the area comprises one or more plants already depicted in the digital image.
4. The method of claim 3 , wherein the plurality of three-dimensional plants simulate weeds targeted for remediation and the one or more plants already depicted in the digital image comprise crops targeted for harvest.
5. The method of claim 3 , wherein the plurality of three-dimensional plants simulate crops targeted for harvest and the one or more plants already depicted in the digital image comprise weeds targeted for remediation.
6. The method of claim 1 , wherein the feature of the area comprises a plurality of environmental conditions that existed in the area during a time interval preceding capture of the digital image that captures the area.
7. The method of claim 1 , further comprising analyzing the digital image to identify the feature of the area.
8. The method of claim 1 , further comprising using the synthetic training image to train a machine learning model to segment or detect, in digital images, plants of a same type as at least some of the three-dimensional synthetic plants.
9. A system for generating a synthetic training image, the system comprising one or more processors and memory storing instructions that, in response to execution of the instructions, cause the one or more processors to:
obtain a digital image that captures an area;
based at least in part on an agricultural history of the area, generate a plurality of three-dimensional synthetic plants; and
generate the synthetic training image to depict the plurality of three-dimensional synthetic plants in the area, wherein the synthetic training image is generated by graphical incorporation of the plurality of three-dimensional synthetic plants with the digital image at a spacing that is determined based at least in part on a feature of the area when the area was captured in the digital image.
10. The system of claim 9 , wherein the feature of the area comprises an environmental condition of the area.
11. The system of claim 10 , wherein the environmental condition of the area comprises one or more plants already depicted in the digital image.
12. The system of claim 11 , wherein the plurality of three-dimensional plants simulate weeds targeted for remediation and the one or more plants already depicted in the digital image comprise crops targeted for harvest.
13. The system of claim 11 , wherein the plurality of three-dimensional plants simulate crops targeted for harvest and the one or more plants already depicted in the digital image comprise weeds targeted for remediation.
14. The system of claim 9 , wherein the feature of the area comprises a plurality of environmental conditions that existed in the area during a time interval preceding capture of the digital image that captures the area.
15. The system of claim 9 , further comprising analyzing the digital image to identify the feature of the area.
16. The system of claim 9 , further comprising using the synthetic training image to train a machine learning model to segment or detect, in digital images, plants of a same type as at least some of the three-dimensional synthetic plants.
17. A non-transitory computer-readable medium for generating a synthetic training image, the medium storing instructions that, in response to execution of the instructions, cause one or more processors to:
obtain a digital image that captures an area;
based at least in part on an agricultural history of the area, generate a plurality of three-dimensional synthetic plants; and
generate the synthetic training image to depict the plurality of three-dimensional synthetic plants in the area, wherein the synthetic training image is generated by graphical incorporation of the plurality of three-dimensional synthetic plants with the digital image at a spacing that is determined based at least in part on a feature of the area when the area was captured in the digital image.
18. The non-transitory computer-readable medium of claim 17 , wherein the feature of the area comprises an environmental condition of the area.
19. The non-transitory computer-readable medium of claim 18 , wherein the environmental condition of the area comprises one or more plants already depicted in the digital image.
20. The non-transitory computer-readable medium of claim 19 , wherein the plurality of three-dimensional plants simulate weeds targeted for remediation and the one or more plants already depicted in the digital image comprise crops targeted for harvest.