Method for recognizing image, computer product and readable storage medium

1. A method for recognizing an image, comprising:

determining categories of a plurality of objects in an image to be detected, and a plurality of object regions where the plurality of objects are located;

determining positions of the object regions in the image to be detected, and sizes of the object regions;

determining a relative positional relationship between the objects according to the positions and the sizes of the object regions; and

obtaining a semantic interpretation of the relative positional relationship between the objects according to the relative positional relationship between the objects;

wherein the determining the relative positional relationship between the objects according to the positions and the sizes of the object regions comprises:

determining a relative positional relationship between objects corresponding to two correlated ones of the object regions according to the positions and the sizes of the two correlated ones of the object regions in a dual-spatial-mask method.

2. The method according to claim 1 , wherein the determining the categories of the plurality of objects in the image to be detected, and the plurality of object regions where the plurality of objects are located comprises:

determining the categories of the plurality of objects in the image to be detected, and the plurality of object regions where the plurality of objects are located using a neutral network.

3. The method according to claim 2 , wherein the neutral network comprises a Regions with Convolutional Neural Network, and the number of region proposals is no less than 200 in the Regions with Convolutional Neural Network.

4. The method according to claim 3 , wherein an intersection over union of region proposals is more than or equal to 0.5.

5. The method according to claim 4 , wherein the intersection over union of region proposals is more than or equal to 0.6 and less than or equal to 0.7.

6. The method according to claim 1 , wherein the method further comprises:

selecting any two of the object regions with their distance within in a preset range as the two correlated ones of the object regions.

7. The method according to claim 1 , wherein the obtaining the semantic interpretation of the relative positional relationship between the objects according to the relative positional relationship between the objects comprises:

inputting the categories of the objects and the relative positional relationship between the objects to a probability knowledge network, and obtaining the semantic interpretation of the relative positional relationship between the objects.

8. A non-transitory readable storage medium, configured to store computer instructions, wherein the computer instructions, when executed by a processor, cause the processor to perform one or more steps in the method according to claim 1 .

9. A computer product, comprising:

a memory configured to store computer instructions; and

one or more processors configured to execute the computer instructions to perform:

determining categories of a plurality of objects in an image to be detected, and a plurality of object regions where the plurality of objects are located;

determining the positions of the object regions in the image to be detected, and sizes of the object regions;

determining a relative positional relationship between the objects according to the positions and the sizes of the object regions; and

obtaining a semantic interpretation of the relative positional relationship between the objects according to the relative positional relationship between the objects;

wherein the one or more processors further configured to execute the computer instructions to perform:

determining a relative positional relationship between objects corresponding to two correlated ones of the object regions according to the positions and the sizes of the two correlated ones of the object regions in a dual-spatial-mask method.

10. The computer product according to claim 9 , wherein the one or more processors further configured to execute the computer instructions to perform:

determining the categories of the plurality of objects in the image to be detected, and the plurality of object regions where the plurality of objects are located using a neutral network.

11. The computer product according to claim 10 , wherein the neutral network comprises a Regions with Convolutional Neutral Network, and the number of region proposals is no less than 200 in the Regions with Convolutional Neutral Network.

12. The computer product according to claim 11 , wherein in the Regions with Convolutional Neutral Network, an intersection over union of region proposals is more than or equal to 0.5.

13. The computer product according to claim 12 , wherein the intersection over union of region proposals is more than or equal to 0.6 and less than or equal to 0.7.

14. The computer product according to claim 9 , wherein the one or more processors further configured to execute the computer instructions to perform:

selecting any two of the object regions with their distance within in a preset range as the two correlated ones of the object regions.

15. The computer product according to claim 9 , wherein the one or more processors further configured to execute the computer instructions to perform:

inputting the categories of the objects and the relative positional relationship between the objects to a probability knowledge network, and obtaining the semantic interpretation of the relative positional relationship between the objects.