Multi-temporal scale analytics
Systems, methods, and computer-readable for multi-temporal scale analysis include obtaining two or more timescales associated with one or more images. A context associated with a monitoring objective is obtained, based on real time analytics or domain specific knowledge. The monitoring objective can include object detection, event detection, pattern recognition, or other. At least a subset of timescales for performing a differential analysis on the one or more images is determined based on the context. Multi timescale surprise detection and clustering are performed using the subset of timescales to determine whether any alerts are to be generated based on entropy based surprises. A set of rules can be created for the monitoring objective based on the differential analytics and alerts or entropy based surprises, if any.
1. A method of analyzing a plurality of images, the method comprising:
obtaining two or more timescales associated with the plurality of images, the plurality of images being captured at different intervals in time corresponding to the two or more timescales;
obtaining a context associated with a monitoring objective, the context providing a baseline pattern for a context based analysis on the plurality of images, the context based analysis including evaluating flagged events with domain specific knowledge, the domain specific knowledge including a deviation from a known schedule;
determining at least a subset of timescales from the two or more timescales based on the context, the subset of timescales comprising one or more timescales for performing the context based analysis on the plurality of images, wherein the one or more timescales comprise one or more first intervals in time from the different intervals in time corresponding to the two or more timescales, the one or more first intervals in time being smaller than one or more second intervals in time corresponding to the two or more timescales, the one or more first intervals in time being associated with the plurality of images captured at or within the one or more first intervals in time;
calculating a stochastic distance between the plurality of images associated with the two or more timescales;
clustering a subset of images of the plurality of images based on the stochastic distance between the plurality of images associated with the two or more timescales, the clustered subset of images of the plurality of images including a corresponding storage size;
generating a score based on the storage size of the clustered subset of images of the plurality of images, the score being a high surprise score when the storage size is greater than a predetermined threshold and the score being a low surprise score when the storage size is less than the predetermined threshold; and
generating one or more alerts based on the context based analysis and the surprise score associated with the clustered subset of images of the plurality of images.
2. The method of claim 1 , wherein the monitoring objective comprises one or more of object detection, event detection, or pattern recognition.
3. The method of claim 1 , wherein the context comprises one or more of a domain specific knowledge or real time analytics.
4. The method of claim 1 , wherein the context based analysis comprises differential analytics on the plurality of images associated with the subset of timescales.
5. The method of claim 4 , further comprising:
generating a set of rules for the monitoring objective, the set of rules being based on at least one of the differential analytics and the one or more alerts.
6. The method of claim 5 , wherein the differential analytics includes a deep learning model, wherein one or more ground truths for the deep learning model are based on the set of rules and the context.
7. The method of claim 6 , wherein obtaining the two or more timescales associated with the plurality of images comprises obtaining the two or more timescales from two or more time series databases associated with the plurality of images.
8. The method of claim 1 , wherein the one or more alerts comprise an entropy based surprise.
9. A system, comprising:
one or more processors; and
a non-transitory computer-readable storage medium containing instructions which, when executed on the one or more processors, cause the one or more processors to perform operations including:
obtaining two or more timescales associated with a plurality of images, the plurality of images being captured at different intervals in time corresponding to the two or more timescales;
obtaining a context associated with a monitoring objective, the context providing a baseline pattern for a context based analysis on the plurality of images, the context based analysis including evaluating flagged events with domain specific knowledge, the domain specific knowledge including a deviation from a known schedule;
determining at least a subset of timescales from the two or more timescales based on the context, the subset of timescales comprising one or more timescales for performing the context based analysis on the plurality of images, wherein the one or more timescales comprise one or more first intervals in time from the different intervals in time corresponding to the two or more timescales, the one or more first intervals in time being smaller than one or more second intervals in time corresponding to the two or more timescales, the one or more first intervals in time being associated with the plurality of images captured at or within the one or more first intervals in time;
calculating a stochastic distance between the plurality of images associated with the two or more timescales;
clustering a subset of images of the plurality of images based on the stochastic distance between the plurality of images associated with the two or more timescales, the clustered subset of images of the plurality of images including a corresponding storage size;
generating a score based on the storage size of the clustered subset of images of the plurality of images, the score being a high surprise score when the storage size is greater than a predetermined threshold and the score being a low surprise score when the storage size is less than the predetermined threshold; and
generating one or more alerts based on the context based analysis and the surprise score associated with the clustered subset of images of the plurality of images.
10. The system of claim 9 , wherein the monitoring objective comprises one or more of object detection, event detection, or pattern recognition.
11. The system of claim 9 , wherein the context comprises one or more of a domain specific knowledge or real time analytics.
12. The system of claim 9 , wherein the context based analysis comprises differential analytics on the plurality of images associated with the subset of timescales.
13. The system of claim 12 , wherein the operations further comprise:
generating a set of rules for the monitoring objective based on at least one of the differential analytics and the one or more alerts.
14. The system of claim 13 , wherein the differential analytics includes a deep learning model, wherein one or more ground truths for the deep learning model are based on the set of rules and the context.
15. The system of claim 14 , wherein obtaining the two or more timescales associated with the plurality of images comprises obtaining the two or more timescales from two or more time series databases associated with the plurality of images.
16. The system of claim 9 , wherein the one or more alerts comprise an entropy based surprise.
17. A non-transitory machine-readable storage medium, including instructions configured to cause a data processing apparatus to perform operations including:
obtaining two or more timescales associated with a plurality of images;
obtaining a context associated with a monitoring objective, the context providing a baseline pattern for a context based analysis on the plurality of images, the context based analysis including evaluating flagged events with domain specific knowledge, the domain specific knowledge including a deviation from a known schedule;
determining at least a subset of timescales from the two or more timescales based on the context, the subset of timescales comprising one or more timescales for performing the context based analysis on the plurality of images, wherein the one or more timescales comprise one or more first intervals in time from different intervals in time corresponding to the two or more timescales, the one or more first intervals in time being smaller than one or more second intervals in time corresponding to the two or more timescales, the one or more first intervals in time being associated with the plurality of images captured at or within the one or more first intervals in time;
calculating a stochastic distance between the plurality of images associated with the two or more timescales;
clustering a subset of images of the plurality of images based on the stochastic distance between the plurality of images associated with the two or more timescales, the clustered subset of images of the plurality of images including a corresponding storage size;
generating a score based on the storage size of the clustered subset of images of the plurality of images, the score being a high surprise score when the storage size is greater than a predetermined threshold and the score being a low surprise score when the storage size is less than the predetermined threshold; and
generating one or more alerts based on the context based analysis and the surprise score associated with the clustered subset of images of the plurality of images.
18. The non-transitory machine-readable storage medium of claim 17 , wherein the monitoring objective comprises one or more of object detection, event detection, or pattern recognition.
19. The non-transitory machine-readable storage medium of claim 18 , wherein the context comprises one or more of a domain specific knowledge or real time analytics.
20. The non-transitory machine-readable storage medium of claim 19 , wherein the context based analysis comprises differential analytics on the plurality of images associated with the subset of timescales.