Machine learning of context data for social and contextual scene inferences
A system for contextual interpretation of a three-dimensional scene includes an object recognition engine that analyzes scene data collected from the three-dimensional scene to identify at least one object present in the three-dimensional scene. The system further includes a contextual inference engine trained on a context data training set to analyze context of the scene by identifying a potential contextual inference associated in memory with the at least one object identified by the object recognition engine; comparing the scene data to a subset of the context data training set identified as satisfying the potential contextual inference; and outputting scene context information conveying the potential contextual inference responsive to a determination that the scene data and the subset of the context data train set satisfy a predetermined correlation.
1. A system for contextual interpretation of a three-dimensional scene, the system comprising:
memory;
a processor;
an object recognition engine stored in the memory and executable by the processor to:
analyze scene data collected from the three-dimensional scene to identify at least a first object and a second object present in the three-dimensional scene; and
a contextual inference engine trained on a context data training set, the contextual inference engine stored in memory and executable by a processor to:
identify a potential contextual inference associated in memory with a recognized spatial relationship between the first object and the second object;
compare the scene data to a subset of the context data training set identified as satisfying the potential contextual inference; and
output scene context information conveying the potential contextual inference responsive to a determination that the scene data and the subset of the context data training set satisfy a predetermined correlation.
2. The system of claim 1 , wherein the potential contextual inference is a use context inference, the use context inference indicating viability of at least one of the first object and the second object for a potential user-object interaction.
3. The system of claim 1 , wherein the potential contextual inference is a behavioral inference, the behavioral inference indicating a motivation for an action of a living object captured in the scene data.
4. The system of claim 1 , wherein the contextual inference engine is further executable to:
compute a confidence metric quantifying a likelihood that the scene data satisfies a correlation with the subset of the context data training set identified as satisfying the potential contextual inference; and
output the scene context information conveying the potential contextual inference responsive to a determination that the computed confidence metric satisfies a predetermined threshold.
5. The system of claim 1 , wherein context data training set includes instances of context data pre-identified as satisfying or not satisfying each of a plurality of contextual inferences.
6. The system of claim 1 , wherein the system further comprises:
at least one environmental sensor that collects the scene data from the three-dimensional scene.
7. The system of claim 1 , wherein the context data training set includes sound data.
8. One or more computer-readable storage media of a tangible article of manufacture encoding computer-executable instructions for executing on a computer system a computer process for contextual interpretation of a three-dimensional scene, the computer process comprising:
analyzing, with a processor, scene data collected from the three-dimensional scene to identify at least a first object and a second object present in the three-dimensional scene;
identifying a potential contextual inference ssociated in memory with a recognized spatial relationship between the first object and the second object;
comparing the scene data to a subset of a context data training set identified as satisfying the potential contextual inference; and
outputting scene context information conveying the potential contextual inference responsive to a determination that the scene data and the subset of the context data training set satisfy a predetermined correlation.
9. The one or more computer-readable storage media of claim 8 , wherein the context data training set includes instances of context data pre-identified as satisfying or not satisfying each of a plurality of contextual inferences associated with different objects.
10. The one or more computer-readable storage media of claim 8 , wherein the potential contextual inference is a use context inference, the use context inference indicating viability of at least one of the first object and the second object for a potential user-object interaction.
11. The one or more computer-readable storage media of claim 8 , wherein the potential contextual inference is a behavioral inference, the behavioral inference indicating a motivation for an action of a living object captured in the scene data.
12. The one or more computer-readable storage media of claim 8 , wherein comparing the scene data to the subset of the context data training set further comprises:
computing a confidence metric quantifying a likelihood that the scene data satisfies a correlation with the subset of the context data training set identified as satisfying the potential contextual inference; and
outputting the scene context information responsive to a determination that the confidence metric satisfies a predetermined threshold.
13. A device for contextual interpretation of a three-dimensional scene, the device comprising:
memory;
a processor;
an object recognition engine stored in the memory and executable by the processor to analyze scene data collected from the three-dimensional scene to identify at least a first object and a second object present in the three-dimensional scene; and
a contextual inference engine trained on a context data training set, the contextual inference engine stored in memory and executable by a processor to:
identify a potential contextual inference associated in memory with a recognized spatial relationship between the first object and the second object;
compare the scene data to a subset of the context data training set identified as satisfying the potential contextual inference; and
output scene context information conveying the potential contextual inference responsive to a determination that the scene data and the subset of the context data training set satisfy a predetermined correlation.
14. The device of claim 13 , wherein the device further comprises:
at least one environmental sensor that collects the scene data from the three-dimensional scene.
15. The device of claim 13 , wherein the potential contextual inference is a use context inference, the use context inference indicating viability of at least one of the first object and the second object for a potential user-object interaction.
16. The device of claim 13 , wherein the potential contextual inference is a behavioral inference, the behavioral inference indicating a motivation for an action of a living object captured in the scene data.
17. The device of claim 13 , wherein the contextual inference engine is further executable to:
compute a confidence metric quantifying a likelihood that the scene data satisfies a correlation with the subset of the context data training set identified as satisfying the potential contextual inference; and
output the scene context information conveying the potential contextual inference responsive to a determination that the computed confidence metric satisfies a predetermined threshold.
18. The device of claim 17 , wherein the context data training set includes sound data.