Predicting performance and success of large-scale vision algorithms

Patent metadata

Field	Value
Publication number	US-9465994-B1
Application number	US-201514629196-A
Country	US
Kind code	B1
Filing date	Feb 23, 2015
Priority date	Feb 23, 2015
Publication date	Oct 11, 2016
Grant date	Oct 11, 2016

Abstract

Where a plurality of machine learning algorithms is available to process information or data in the furtherance of a task, one of the algorithms may be identified as particularly well-suited or appropriate based on attributes of the information or data. Such attributes of the imaging data may be determined by any means, and a prediction as to the performance (e.g., one or more metrics) or success of each of the algorithms may be made. One of the algorithms may ultimately be selected based on such predictions, as well as the computing resources that are available for executing the algorithms, and any other relevant constraints.

First claim

What is claimed is:

1. An imaging device comprising:
an imaging sensor;
at least one memory device; and
at least one computer processor,
wherein the at least one computer processor is configured to implement one or more services, and
wherein the one or more services are configured to:
capture imaging data using the imaging sensor;
determine at least one spectral property of the imaging data;
identify a plurality of algorithms for performing a discrete task using the at least one computer processor;
predict at least one performance metric of each of the plurality of algorithms for performing the discrete task on the imaging data based at least in part on the at least one spectral property of the imaging data;
select one of the plurality of algorithms for performing the discrete task on the imaging data based at least in part on the at least one predicted performance metric;
execute the selected algorithm for performing the discrete task on the imaging data using the at least one computer processor.

2. The imaging device of claim 1 , wherein the one or more services are further configured to:
determine at least one label of at least one scene expressed in the imaging data,
wherein the at least one performance metric of each of the plurality of algorithms is predicted based at least in part on the at least one label of the at least one scene expressed in the imaging data.

3. The imaging device of claim 1 , wherein the at least one performance metric is at least one of:
an expected amount of memory consumed by each of the plurality of algorithms for performing the discrete task on the imaging data;
an expected level of accuracy or precision of output of each of the plurality of algorithms for performing the discrete task on the imaging data;
an expected rate of success of each of the plurality of algorithms for performing the discrete task on the imaging data; or
an expected time of execution of each of the plurality of algorithms for performing the discrete task on the imaging data.

4. The imaging device of claim 1 , wherein the discrete task is at least one of:
character recognition; determination of structure from motion; edge detection; image-based three-dimensional modeling; image combination; image compression; image conversion; image correction; image filtering; image measurement; image modeling; image noise reduction; image quantization; image sampling; image scaling; image segmentation; image sharpening; image smoothing; image transformation; image zooming; imaging device pose estimation; imaging device-intrinsic parameter determination; object classification; object detection; object recognition; object segmentation; or object tracking.

5. A computer-implemented method comprising:
identifying imaging data captured using at least one imaging device, wherein the imaging data is stored in at least one data store;
determining at least one spectral property of the imaging data using at least one computer processor;
identifying a plurality of algorithms for performing a predetermined task on the imaging data;
predicting at least one performance metric of at least one of the plurality of algorithms for performing the predetermined task on the imaging data based at least in part on the at least one spectral property using the at least one computer processor; and
selecting one of the plurality of algorithms for performing the predetermined task on the imaging data based at least in part on at least one of the at least one spectral property of the imaging data or the at least one predicted performance metric using the at least one computer processor.

6. The computer-implemented method of claim 5 , further comprising:
determining at least one resource accessible by the at least one computer processor,
wherein the at least one algorithm for performing the predetermined task on the imaging data is selected based at least in part on the at least one resource.

7. The computer-implemented method of claim 5 , further comprising:
executing the at least one algorithm for performing the predetermined task on the imaging data using the at least one computer processor.

8. The computer-implemented method of claim 5 , wherein the at least one performance metric is at least one of:
an amount of memory required to execute the at least one of the plurality of algorithms using the at least one computer processor;
a level of accuracy or precision of output of the at least one of the plurality of algorithms;
a rate of success of the at least one of the plurality of algorithms; or
a time required to execute the at least one of the plurality of algorithms using the at least one computer processor.

9. The computer-implemented method of claim 5 , further comprising:
identifying at least one background feature expressed in the imaging data; and
determining the at least one spectral property of the imaging data based at least in part on the at least one background feature.

10. The computer-implemented method of claim 5 , wherein the at least one spectral property is at least one of:
at least one color expressed in the imaging data;
at least one dimension associated with the imaging data;
at least one representation of at least one horizontal texture or edge component;
at least one representation of at least one vertical texture or edge component;
at least one representation of at least one directional texture or edge component;
an indicator of chromaticity of the imaging data;
an indicator of color saturation of the imaging data;
an indicator of at least one pixel feature in at least one scale or level of resolution;
an indicator of pixel heterogeneity; or
an indicator of pixel homogeneity.

11. The computer-implemented method of claim 5 , wherein determining the at least one spectral property of the imaging data further comprises:
providing at least some of the imaging data as an input to at least one filtering algorithm;
processing the at least some of the imaging data according to the at least one filtering algorithm using the at least one computer processor; and
determining the at least one spectral property of the imaging data based at least in part on an output of the at least one filtering algorithm using the at least one computer processor.

12. The computer-implemented method of claim 11 , wherein the at least one filtering algorithm is one of a fast Fourier transform or a discrete Fourier transform.

13. The computer-implemented method of claim 5 , further comprising:
determining at least one scene tag associated with the imaging data based at least in part on the at least one spectral property of the imaging data using the at least one computer processor; and
selecting the one of the plurality of algorithms for performing the predetermined task on the imaging data based at least in part on the at least one scene tag.

14. The computer-implemented method of claim 13 , wherein selecting the at least one algorithm for performing the predetermined task on the imaging data based at least in part on the at least one scene tag further comprises:
providing the at least one scene tag to a classifier as an input; and
selecting the at least one algorithm for performing the predetermined task on the imaging data based at least in part on an output of the classifier.

15. The computer-implemented method of claim 5 , further comprising:
identifying metadata associated with the imaging data using the at least one computer processor; and
selecting the one of the plurality of algorithms for performing the predetermined task on the imaging data based at least in part on the m

Assignees

• Amazon Tech Inc

Inventors

• Mishra Pragyana K

Classifications

• H04N5/77 Primary

  between a recording apparatus and a television camera · CPC title

• G06V10/87

  using selection of the recognition techniques, e.g. of a classifier in a multiple classifier system · CPC title

• H04N7/181

  for receiving images from a plurality of remote sources · CPC title

• G06F18/285

  Selection of pattern recognition techniques, e.g. of classifiers in a multi-classifier system · CPC title

• G06T2207/20056

  Discrete and fast Fourier transform, [DFT, FFT] · CPC title