Multi-sync ensemble model for device localization

1. A method for determining a location of a mobile device, the method comprising:

collecting image data captured by a camera on the mobile device;

analyzing the image data to assign a score to each of a plurality of localization models, the score representing a predicted accuracy of a corresponding localization model in determining the location of the mobile device;

selecting a localization model from the plurality of localization models based on the score assigned to each of the plurality of localization models;

providing the image data as input to the selected localization model, the selected localization model identifying a feature included in the input image data that matches a feature in the selected localization model, and outputting a candidate location based on the matching feature in the selected localization model; and

determining the location of the mobile device based on the candidate location.

2. The method of claim 1 , wherein analyzing the image data comprises applying a pre-configured decision tree to the image data.

3. The method of claim 1 , wherein analyzing the image data comprises applying a reinforcement learning model to the image data, the reinforcement learning model using the predicted accuracies of the plurality of localization models as an expected reward.

4. The method of claim 1 , wherein analyzing the image data comprises:

determining an illumination level of the image data; and

assigning the score to each of the plurality of localization models based on the illumination level.

5. The method of claim 1 , wherein analyzing the image data comprises:

determining, based on the image data, whether the image data represents an indoor environment or an outdoor environment; and

assigning the score to each of the plurality of localization models based on the determination.

6. The method of claim 5 , wherein determining whether the image data represents the indoor environment or the outdoor environment comprises:

identifying one or more features from the image data, the one or more features including a physical object captured in the image data; and

determining, based on the one or more identified features, whether the image data represents the indoor environment or the outdoor environment.

7. The method of claim 6 , wherein determining, based on the one or more identified features, whether the image data represents the indoor environment or the outdoor environment comprises characterizing the physical object as an indoor object or an outdoor object, wherein the determination is based on the characterization of the physical object.

8. The method of claim 1 , wherein the plurality of localization models include two or more of: a point cloud based model, a plane matching model, a line matching model, a geographic information system (GIS) model, a building recognition model, an object recognition model, a semantic matching model, a cube matching model, a cylinder matching model, a horizon matching model, a light source matching model, and a landscape recognition model.

9. The method of claim 1 , further comprising:

selecting one or more additional localization models from the plurality of localization models based on the score assigned to each of the plurality of localization models; and

providing the image data as input to the one or more additional localization models, the one or more additional localization models outputting one or more additional candidate locations,

wherein determining the location of the mobile device is further based on the one or more additional candidate locations.

10. The method of claim 9 , wherein determining the location of the mobile device comprises determining an average location of the candidate location and the one or more additional candidate locations.

11. A non-transitory computer-readable storage medium comprising instructions that, when executed by a computing device, cause the computing device to perform operations including:

collecting image data captured by a camera on a mobile device;

analyzing the image data to assign a score to each of a plurality of localization models, the score representing a predicted accuracy of a corresponding localization model in determining a location of the mobile device;

selecting a localization model from the plurality of localization models based on the score assigned to each of the plurality of localization models;

providing the image data as input to the selected localization model, the selected localization model identifying a feature in the input image data that matches a feature in the selected localization model, and outputting a candidate location based on the matching feature in the selected localization model; and

determining the location of the mobile device based on the candidate location.

12. The computer-readable storage medium of claim 11 , wherein analyzing the image data comprises applying a pre-configured decision tree to the image data.

13. The computer-readable storage medium of claim 11 , wherein analyzing the image data comprises applying a reinforcement learning model to the image data, the reinforcement learning model using the predicted accuracies of the plurality of localization models as an expected reward.

14. The computer-readable storage medium of claim 11 , wherein analyzing the image data comprises:

determining an illumination level of the image data; and

assigning the score to each of the plurality of localization models based on the illumination level.

15. The computer-readable storage medium of claim 11 , wherein analyzing the image data comprises:

determining, based on the image data, whether the image data represents an indoor environment or an outdoor environment; and

assigning the score to each of the plurality of localization models based on the determination.

16. The computer-readable storage medium of claim 15 , wherein determining whether the image data represents the indoor environment or the outdoor environment comprises:

identifying one or more features from the image data, the one or more features including a physical object captured in the image data; and

determining, based on the one or more identified features, whether the image data represents the indoor environment or the outdoor environment.

17. The computer-readable storage medium of claim 16 , wherein determining, based on the one or more identified features, whether the image data represents the indoor environment or the outdoor environment comprises characterizing the physical object as an indoor object or an outdoor object, wherein the determination is based on the characterization of the physical object.

18. The computer-readable storage medium of claim 11 , wherein the plurality of localization models include two or more of: a point cloud based model, a plane matching model, a line matching model, a geographic information system (GIS) model, a building recognition model, an object recognition model, a semantic matching model, a cube matching model, a cylinder matching model, a horizon matching model, a light source matching model, and a landscape recognition model.

19. The computer-readable storage medium of claim 11 , wherein the operations further include:

selecting one or more additional localization models from the plurality of localization models based on the score assigned to each of the plurality of localization models; and

providing the image data as input to the one or more additional localization models, the one or more additional localization models outputting one or more additional candidate locations,

wherein determining the location of the mobile device is further based on the one or more additional candidate locations.

20. A mobile device comprising:

a camera configured to capture image data;

a data store storing a plurality of localization models; and

a localization subsystem configured to:

analyze the image data to assign a score to each of the plurality of localization models, the score representing a predicted accuracy of a corresponding localization model in determining the location of the mobile device;

select a localization model from the plurality of localization models based on the score assigned to each of the plurality of localization models;

provide the image data as input to the selected localization model, the selected localization model identifying a feature included in the input image data that matches a feature in the selected localization model, and outputting a candidate location based on the matching feature in the selected localization model; and

determine the location of the mobile device based on the candidate location.