Image processing system including training model based upon iteratively biased loss function and related methods
An image processing system may include a processor and an associated memory configured to store training data that includes training geospatial images. The processor may be configured to operate a training model to identify a given feature from each of the training geospatial images, and to iteratively apply a bias factor to a loss function based upon a number of incorrectly identified pixels for the given feature. The bias factor may be reduced with each iteration.
1. An image processing system comprising:
a processor and an associated memory configured to
store training data comprising a plurality of training geospatial images, and
operate a training model to identify a given feature from each of the plurality of training geospatial images, and to iteratively apply a bias factor to a loss function based upon a number of incorrectly identified pixels for the given feature, the bias factor being reduced with each iteration.
2. The image processing system of claim 1 wherein the training model comprises a mask-based training model.
3. The image processing system of claim 2 wherein each of the plurality of training geospatial images comprises a plurality of pixels; and wherein said processor is configured to iteratively apply the bias factor for masked ones of the plurality of pixels.
4. The image processing system of claim 1 wherein the bias factor comprises an exponential function.
5. The image processing system of claim 4 wherein the bias factor comprises a summation of the exponential function.
6. The image processing system of claim 5 wherein the summation of the exponential function is multiplied by the loss function.
7. The image processing system of claim 1 wherein said processor is configured to operate the training model to identify the given feature from a plurality of geospatial images.
8. A method of processing an image comprising:
using a processor and an associated memory to
store training data comprising a plurality of training geospatial images, and
operate a training model to identify a given feature from each of the plurality of training geospatial images, and to iteratively apply a bias factor to a loss function based upon a number of incorrectly identified pixels for the given feature, the bias factor being reduced with each iteration.
9. The method of claim 8 wherein the training model comprises a mask-based training model.
10. The method of claim 9 wherein each of the plurality of training geospatial images comprises a plurality of pixels; and wherein using the processor comprises using the processor to iteratively apply the bias factor for masked ones of the plurality of pixels.
11. The method of claim 8 wherein the bias factor comprises an exponential function.
12. The method of claim 11 wherein the bias factor comprises a summation of the exponential function.
13. The method of claim 12 wherein the summation of the exponential function is multiplied by the loss function.
14. The method of claim 8 wherein using the processor comprises using the processor to operate the training model to identify the given feature from a plurality of geospatial images.
15. A non-transitory computer readable medium for processing an image, the non-transitory computer readable medium comprising computer executable instructions that when executed by a processor cause the processor to perform operations comprising:
storing training data comprising a plurality of training geospatial images; and
operating a training model to identify a given feature from each of the plurality of training geospatial images, and to iteratively apply a bias factor to a loss function based upon a number of incorrectly identified pixels for the given feature, the bias factor being reduced with each iteration.
16. The non-transitory computer readable medium of claim 15 wherein the training model comprises a mask-based training model.
17. The non-transitory computer readable medium of claim 16 wherein each of the plurality of training geospatial images comprises a plurality of pixels; and wherein the operations comprise operating the training model to iteratively apply the bias factor for masked ones of the plurality of pixels.
18. The non-transitory computer readable medium of claim 15 wherein the bias factor comprises an exponential function.
19. The non-transitory computer readable medium of claim 18 wherein the bias factor comprises a summation of the exponential function.
20. The non-transitory computer readable medium of claim 19 wherein the summation of the exponential function is multiplied by the loss function.
21. The non-transitory computer readable medium of claim 15 wherein the operations comprise operating the training model to identify the given feature from a plurality of geospatial images.