System and method using deep learning machine vision to analyze localities
A system, method, and computer-readable storage medium are disclosed that execute machine vision operations to categorize a locality. At least one embodiment accesses a map image of a locality, where the map image includes geographical artefacts corresponding to entities within the locality; analyzes the map image to detect the entities in the locality using the geographical artefacts; assigns entity classes to detected entities in the locality; assigns a locality score to the locality based on entity classes included in the locality; retrieves street view images for one or more of the detected entities in the locality; and analyzes street view images of the detected entities to assign one or more further classifications to the detected entities. Other embodiments include corresponding computer systems, apparatus, and computer programs recorded on one or more computer storage devices, each configured to perform the actions of the method.
1. A computer-implemented method for using machine vision, the method comprising:
training a neural network to extract regions of segmented pixel areas of a map image, the neural network comprising a first convolutional neural network and a second convolutional network, the training comprising classifying entities from a plurality of entities based upon at least one of text and icons represented within the segmented pixel areas, the classifying entities comprising classifying entities into a first type and a second type, the first convolutional neural network being trained using entities of the first type and the second convolutional neural network being trained using entities of the second type;
accessing a map image of a locality, wherein the map image of the locality includes geographical artefacts corresponding to entities within the locality;
analyzing, using the neural network, the map image to detect entities in the locality using the geographical artefacts;
assigning entity classes to the detected entities in the locality, the assigning the entities in the locality comprising assigning detected entities in the locality to the first type or the second type;
assigning a locality score to the locality based on the entity classes included in the locality;
retrieving street view images for one or more of the detected entities in the locality; and analyzing the street view images of the detected entities to assign one or more classifications to the detected entities, certain characteristics of the street view images being used to assign the one or more classifications, the one or more classifications being used in combination with the locality score of the locality to classify an entity at the locality.
2. The computer-implemented method of claim 1 , wherein:
the geographical artefacts include one or more of text and/or icons associated with the entities within the locality.
3. The computer-implemented method of claim 1 , wherein analyzing the map image comprises:
providing the map image to the neural network, wherein the neural network is configured to separate portions of the map image having geographical artefacts from portions of the map image that do not have geographical artefacts;
reconstructing the map image using the portions of the map image having geographical artefacts; and
identifying entities in the reconstructed map image using geographical artefacts of the reconstructed map image.
4. The computer-implemented method of claim 3 , further comprising:
accessing a database for the locality to identify the entities using textual portions of the reconstructed map image.
5. The computer-implemented method of claim 1 , wherein assigning the locality score comprises:
clustering entities having similar entity classes within the locality into cluster types; and
assigning the locality score to the locality based on the clustered entities within the locality.
6. The computer-implemented method of claim 5 , further comprising:
assigning the locality scores to the locality using a percentage based analysis, wherein a percentage is assigned to a cluster type based on a number of the entities included in the cluster type to a total number of clustered entities.
7. The computer-implemented method of claim 1 , wherein
analyzing the map image includes
providing the map image to the neural network, wherein the neural network is configured to separate portions of the map image having geographical artefacts from portions of the map image that do not have geographical artefacts;
reconstructing the map image using the portions of the map image having geographical artefacts;
detecting entities in the reconstructed map using the geographical artefacts;
identifying entity classes of the detected entities;
wherein assigning the locality score includes
clustering entities with similar entity classes within the locality, wherein a cluster has a cluster type corresponding to the entity classes within the cluster; and
wherein analyzing the street view images includes
providing street views of the identified entities to an input of a further neural network, wherein the further neural network has been trained to classify the entities within a cluster based on the cluster type.
8. A system comprising:
a processor;
a data bus coupled to the processor;
and a non-transitory, computer-readable storage medium embodying computer program code, the non-transitory, computer-readable storage medium being coupled to the data bus, the computer program code interacting with a plurality of computer operations and comprising instructions executable by the processor and configured for:
training a neural network to extract regions of segmented pixel areas of a map image, the neural network comprising a first convolutional neural network and a second convolutional network, the training comprising classifying entities from a plurality of entities based upon at least one of text and icons represented within the segmented pixel areas, the classifying entities comprising classifying entities into a first type and a second type, the first convolutional neural network being trained using entities of the first type and the second convolutional neural network being trained using entities of the second type;
accessing a map image of a locality, wherein the map image of the locality includes geographical artefacts corresponding to entities within the locality;
analyzing, using the neural network, the map image to detect entities in the locality using the geographical artefacts;
assigning entity classes to the detected entities in the locality, the assigning the entities in the locality comprising assigning detected entities in the locality to the first type or the second type;
assigning a locality score to the locality based on the entity classes included in the locality;
retrieving street view images for one or more of the detected entities in the locality; and
analyzing the street view images of the detected entities to assign one or more classifications to the detected entities, certain characteristics of the street view images being used to assign the one or more classifications, the one or more classifications being used in combination with the locality score of the locality to classify an entity at the locality.
9. The system of claim 8 , wherein:
the geographical artefacts include one or more of text and/or icons associated with the entities within the locality.
10. The system of claim 8 , wherein analyzing the map image comprises:
providing the map image to the neural network, wherein the neural network is configured to separate portions of the map image having geographical artefacts from portions of the map image that do not have geographical artefacts;
reconstructing the map image using the portions of the map image having geographical artefacts; and
identifying entities in the reconstructed map image using geographical artefacts of the reconstructed map image.
11. The system of claim 10 , further comprising:
accessing a database for the locality to identify the entities using textual portions of the reconstructed map image.
12. The system of claim 8 , wherein assigning the locality score comprises:
clustering entities having similar entity classes within the locality into cluster types; and
assigning the locality score to the locality based on the clustered entities within the locality.
13. The system of claim 12 , further comprising:
assigning the locality scores to the locality using a percentage based analysis, wherein a percentage is assigned to a cluster based on a number of the entities included in the cluster to a total number of clustered entities.
14. The system of claim 8 , wherein:
analyzing the map image includes
providing the map image to the neural network, wherein the neural network is configured to separate portions of the map image having geographical artefacts from portions of the map image that do not have geographical artefacts;
reconstructing the map image using the portions of the map image having geographical artefacts;
detecting entities in the reconstructed map using the geographical artefacts;
identifying entity classes of the detected entities;
wherein assigning the locality score includes
clustering entities with similar entity classes within the locality, wherein a cluster has a cluster type corresponding to the entity classes within the cluster; and
wherein analyzing the street view images includes
providing street views of the identified entities to an input of a further neural network, wherein the further neural network has been trained to classify entities within a cluster based on the cluster type.
15. A non-transitory, computer-readable storage medium embodying computer program code, the computer program code comprising computer executable instructions configured for:
training a neural network to extract regions of segmented pixel areas of a map image, the neural network comprising a first convolutional neural network and a second convolutional network, the training comprising classifying entities from a plurality of entities based upon at least one of text and icons represented within the segmented pixel areas, the classifying entities comprising classifying entities into a first type and a second type, the first convolutional neural network being trained using entities of the first type and the second convolutional neural network being trained using entities of the second type;
accessing a map image of a locality, wherein the map image of the locality includes geographical artefacts corresponding to entities within the locality;
analyzing, using the neural network, the map image to detect entities in the locality using the geographical artefacts;
assigning entity classes to the detected entities in the locality, the assigning the entities in the locality comprising assigning detected entities in the locality to the first type or the second type;
assigning a locality score to the locality based on the entity classes included in the locality;
retrieving street view images for one or more of the detected entities in the locality; and
analyzing the street view images of the detected entities to assign one or more classifications to the detected entities, certain characteristics of the street view images being used to assign the one or more classifications, the one or more classifications being used in combination with the locality score of the locality to classify an entity at the locality.
16. The non-transitory, computer-readable storage medium of claim 15 , wherein:
the geographical artefacts include one or more of text and/or icons associated with the entities within the locality.
17. The non-transitory, computer-readable storage medium of claim 15 , wherein analyzing the map image comprises:
providing the map image to the neural network, wherein the neural network is configured to separate portions of the map image having geographical artefacts from portions of the map image that do not have geographical artefacts;
reconstructing the map image using the portions of the map image having geographical artefacts; and
identifying entities in the reconstructed map image using geographical artefacts of the reconstructed map image.
18. The non-transitory, computer-readable storage medium of claim 17 , further comprising:
accessing a database for the locality to identify the entities using textual portions of the reconstructed map image.
19. The non-transitory, computer-readable storage medium of claim 15 , wherein assigning the locality score comprises:
clustering entities having similar entity classes within the locality into cluster types; and
assigning the locality score to the locality based on the clustered entities within the locality.
20. The non-transitory, computer-readable storage medium of claim 19 , further comprising:
assigning the locality scores to the locality using a percentage based analysis, wherein a percentage is assigned to a cluster based on a number of the entities included in the cluster in relation to a total number of clustered entities.