Adapting pre-trained classification algorithms

1. A computer-enabled method for analyzing image data, the method comprising:

receiving an input image, wherein the input image is of a second data domain;

providing the input image to a preprocessing algorithm to obtain a transformed image, wherein the preprocessing algorithm is trained to transform data of the second data domain to data of a first data domain;

providing the transformed image to a trained algorithm to analyze the transformed image, wherein the trained algorithm is trained based on training data of the first data domain.

2. The method of claim 1 ,

wherein data of the first data domain is associated with a first image style,

wherein data of the second data domain is associated with a second image style, and

wherein the preprocessing algorithm is trained to transform a first image of the first image style to a second image of the second image style while preserving content of the first image in the second image.

3. The method of claim 1 ,

wherein data of the first data domain is associated with a first image resolution range, and

wherein data of the second data domain is associated with a second image resolution range.

4. The method of claim 1 , further comprising training the preprocessing algorithm, wherein training the preprocessing algorithm comprises:

receiving a first set of training data of the first data domain;

receiving a second set of training data of the second data domain;

training a first image generator, based on the first set of training data and the second set of training data, to transform a given image of the first data domain to an image of the second data domain; and

training a second image generator, based on the first set of training data and the second set of training data, to transform a given image of the second data domain to an image of the first data domain, wherein the preprocessing algorithm comprises the second image generator.

5. The method of claim 4 , wherein the first image generator comprises a first convolution neural network, and the second image generator comprises a second convolution neural network.

6. The method of claim 5 , further comprising optimizing the second image generator by performing normalization across a plurality of channels in the second convolution neural network,

wherein the plurality of channels represent outputs of a convolution layer, and

wherein the plurality of channels correspond to a single pixel.

7. The method of claim 4 , further comprising:

providing an original image of the second set of training data to the second image generator to obtain a corresponding processed image; and

calculating an adversarial loss based on the original image and the corresponding processed image.

8. The method of claim 7 , further comprising:

providing the corresponding processed image to the first image generator to obtain a corresponding cycled image; and

calculating a cycle loss based on the original image and the corresponding cycled image.

9. The method of claim 7 , wherein the original image is a portion of an image in the second set of training data, wherein the portion is selected based on one or more annotations associated with the image in the second set of training data.

10. The method of claim 9 , wherein the one or more annotations represent one or more objects of interest.

11. The method of claim 4 , wherein the first image generator is configured to receive an input image of a first size and output an output image of a second size, wherein the second size is smaller than the first size.

12. The method of claim 4 , wherein the second image generator is configured to receive an input image of the second size and output an output image of the first size.

13. The method of claim 4 , wherein the second convolution neural network comprises a skip connection associated with two layers of the neural network, wherein the two layers are associated with a same spatial resolution.

14. The method of claim 4 , wherein training the second image generator comprises calculating an identity mapping value.

15. The method of claim 4 , wherein the first image generator and the second image generator are part of a modified CycleGAN implementation.

16. The method of claim 1 , wherein data of the second data domain is captured by one or more unmanned aircraft vehicles.

17. The method of claim 1 , wherein data of the first data domain is associated with the ImageNet database.

18. The method of claim 1 , wherein the input image is a first input image, wherein the transformed image is a first transformed image, and wherein the preprocessing algorithm is a first preprocessing algorithm, further comprising:

receiving a second input image, wherein the second input image is of a third data domain;

providing the second input image to a second preprocessing algorithm to obtain a second transformed image, wherein the second preprocessing algorithm is trained to transform data of the third data domain to data of the first data domain;

providing the second transformed image to the trained algorithm to analyze the second transformed image.

19. An electronic device, comprising:

one or more processors;

a memory; and

one or more programs, wherein the one or more programs are stored in the memory and configured to be executed by the one or more processors, the one or more programs including instructions for:

receiving an input image, wherein the input image is of a second data domain;

providing the input image to a preprocessing algorithm to obtain a transformed image, wherein the preprocessing algorithm is trained to transform data of the second data domain to data of a first data domain;

providing the transformed image to a trained algorithm to analyze the transformed image, wherein the trained algorithm is trained based on training data of the first data domain.

20. A non-transitory computer-readable storage medium storing one or more programs, the one or more programs comprising instructions, which when executed by one or more processors of an electronic device having a display, cause the electronic device to:

receive an input image, wherein the input image is of a second data domain;

provide the input image to a preprocessing algorithm to obtain a transformed image, wherein the preprocessing algorithm is trained to transform data of the second data domain to data of a first data domain;

provide the transformed image to a trained algorithm to analyze the transformed image, wherein the trained algorithm is trained based on training data of the first data domain.