Determining a post-processing configuration for post-processing output data from a pipeline

1. A method, comprising:

configuring a pipeline for a target device, the pipeline including a signal processing component and a machine learning component, wherein the pipeline is configured to receive input data and generate output data based on the input data; and

determining multiple post-processing configurations for post-processing the output data, wherein a post-processing configuration is configured to generate a detectable event based on the output data, and wherein the multiple post-processing configurations are generated using a multi-objective optimization that varies one or more parameters for generating the detectable event.

2. The method of claim 1 , further comprising:

receiving a selection of a first post-processing configuration from among the multiple post-processing configurations; and

configuring the first post-processing configuration for an embedded device corresponding to the target device.

3. The method of claim 1 , wherein the one or more parameters are configured to control filtering the output data by changing a window that averages detections in the output data.

4. The method of claim 1 , wherein the one or more parameters are configured to control thresholding of the output data by changing a level at which detections in the output data are determined.

5. The method of claim 1 , wherein the multi-objective optimization determines the multiple post-processing configurations by determining optimizations of precision and recall.

6. The method of claim 1 , wherein the multiple post-processing configurations represent a Pareto frontier that based on precision and recall.

7. The method of claim 1 , wherein the multiple post-processing configurations represent a Pareto frontier, the method further comprising:

configuring a graphical user interface (GUI) to display the multiple post-processing configurations as multiple points of the Pareto frontier.

8. The method of claim 1 , further comprising:

configuring a GUI to display a first post-processing configuration representing a default post-processing configuration before applying the multi-objective optimization.

9. The method of claim 1 , further comprising:

configuring the input data in a background stream and adding one or more detectable events at predetermined locations in the background stream.

10. The method of claim 1 , further comprising:

configuring the input data in a background stream and adding one or more anomalies at predetermined locations in the background stream.

11. The method of claim 1 , further comprising:

determining a latency of an embedded device corresponding to the target device; and

using the latency to change the one or more parameters.

12. An apparatus, comprising:

a memory; and

a processor configured to execute instructions stored in the memory to:

receive an input indicating a target device;

configure a pipeline for the target device, the pipeline including a signal processing component and a machine learning component, wherein the pipeline is configured to receive input data and generate output data based on the input data; and

determine multiple post-processing configurations for post-processing the output data, wherein a post-processing configuration is configured to generate a detectable event based on the output data, and wherein the multiple post-processing configurations are generated using a multi-objective optimization that varies one or more parameters for generating the detectable event.

13. The apparatus of claim 12 , wherein the processor is further configured to execute instructions stored in the memory to:

receive a selection of a first post-processing configuration from among the multiple post-processing configurations; and

configure the first post-processing configuration for an microcontroller corresponding to the target device.

14. The apparatus of claim 12 , wherein the pipeline implements a sound classifier.

15. The apparatus of claim 12 , wherein the multi-objective optimization determines the multiple post-processing configurations by determining optimizations of a false acceptance rate and a false rejection rate.

16. The apparatus of claim 12 , wherein the multiple post-processing configurations represent a Pareto frontier that based on a false acceptance rate and a false rejection rate.

17. A non-transitory computer readable medium storing instructions operable to cause one or more processors to perform operations comprising:

configuring a pipeline for a target device, the pipeline including a signal processing component and a machine learning component, wherein the pipeline is configured to receive input data and generate output data based on the input data; and

determining multiple post-processing configurations for post-processing the output data, wherein a post-processing configuration is configured to generate a detectable event based on the output data, and wherein the multiple post-processing configurations are generated using a multi-objective optimization that varies one or more parameters for generating the detectable event.

18. The non-transitory computer readable medium storing instructions of claim 17 , the operations further comprising:

receiving a selection of a first post-processing configuration from among the multiple post-processing configurations; and

configuring the first post-processing configuration for an embedded device corresponding to the target device.

19. The non-transitory computer readable medium storing instructions of claim 17 , wherein the input data in the first stream is audio data from a microphone.

20. The non-transitory computer readable medium storing instructions of claim 17 , the operations further comprising:

configuring the input data in a background stream and adding one or more anomalies at predetermined locations in the background stream.