Optimal data sampling for image analysis

1. A method comprising:

receiving image data with a first image resolution;

estimating accuracy of image analysis on the image data with the first image resolution;

determining an optimal image resolution for sampling the image data based on the estimated accuracy of image analysis and a learned model trained to determine different optimal image resolutions for different images with different image resolutions, wherein the optimal image resolution is lower than the first image resolution;

sampling the image data at the optimal image resolution; and

performing image analysis on the sampled image data resulting from the sampling.

2. The method of claim 1 , wherein the optimal image resolution represents an optimal trade-off between the estimated accuracy of image analysis and one or more costs associated with the image analysis, and the optimal image resolution satisfies at least one of a target accuracy or a target cost.

3. The method of claim 2 , wherein the learned model is trained based on an accuracy model for estimating accuracy of image analysis and one or more cost models for estimating the one or more costs associated with the image analysis.

4. The method of claim 3 , wherein the accuracy model is one of a concave function or a monotonically increasing function.

5. The method of claim 3 , wherein the one or more cost models comprise a linear model fitted via regressions using training data.

6. The method of claim 3 , wherein the one or more cost models estimate at least one of the following costs associated with the image analysis: execution time, memory consumption, central processing unit (CPU) consumption, input/output (I/O) bandwidth consumption, or network bandwidth consumption.

7. The method of claim 2 , wherein the sampling the image data at the optimal image resolution comprises down-sampling different image dimensions of the image data by different factors.

8. The method of claim 1 , further comprising:

storing the sampled image data instead of the image data on a storage device, thereby reducing data storage consumption.

9. The method of claim 1 , wherein the performing the image analysis on the sampled image data instead of the image data reduces computational time and resource consumption.

10. A system, comprising:

at least one processor; and

a non-transitory processor-readable memory device storing instructions that when executed by the at least one processor causes the at least one processor to perform operations including:

receiving image data with a first image resolution;

estimating accuracy of image analysis on the image data with the first image resolution;

determining an optimal image resolution for sampling the image data based on the estimated accuracy of image analysis and a learned model trained to determine different optimal image resolutions for different images with different image resolutions, wherein the optimal image resolution is lower than the first image resolution;

sampling the image data at the optimal image resolution; and

performing image analysis on the sampled image data resulting from the sampling.

11. The system of claim 10 , wherein the optimal image resolution represents an optimal trade-off between the estimated accuracy of image analysis and one or more costs associated with the image analysis, and the optimal image resolution satisfies at least one of a target accuracy or a target cost.

12. The system of claim 11 , wherein the learned model is trained based on an accuracy model for estimating accuracy of image analysis and one or more cost models for estimating the one or more costs associated with the image analysis.

13. The system of claim 12 , wherein the accuracy model is one of a concave function or a monotonically increasing function.

14. The system of claim 12 , wherein the one or more cost models comprise a linear model fitted via regressions using training data.

15. The system of claim 12 , wherein the one or more cost models estimate at least one of the following costs associated with the image analysis: execution time, memory consumption, central processing unit (CPU) consumption, input/output (I/O) bandwidth consumption, or network bandwidth consumption.

16. The system of claim 11 , wherein the sampling the image data at the optimal image resolution comprises down-sampling different image dimensions of the image data by different factors.

17. The system of claim 10 , wherein the operations further comprise:

storing the sampled image data instead of the image data on a storage device, thereby reducing data storage consumption.

18. The system of claim 10 , wherein the performing the image analysis on the sampled image data instead of the image data reduces computational time and resource consumption.

19. A non-transitory computer readable storage medium including instructions to perform a method comprising:

receiving image data with a first image resolution;

estimating accuracy of image analysis on the image data with the first image resolution;

determining an optimal image resolution for sampling the image data based on the estimated accuracy of image analysis and a learned model trained to determine different optimal image resolutions for different images with different image resolutions, wherein the optimal image resolution is lower than the first image resolution;

sampling the image data at the optimal image resolution; and

performing image analysis on the sampled image data resulting from the sampling.

20. The computer readable storage medium of claim 19 , wherein the optimal image resolution represents an optimal trade-off between the estimated accuracy of image analysis and one or more costs associated with the image analysis.