Method of processing object detection data

1. A computer-implemented method comprising:

receiving, from an object detection system, object detection data comprising a plurality of detection outputs associated with different respective regions of an image, wherein a first detection output of the plurality of detection outputs is associated with a first region of the image and comprises a plurality of received detection characteristics;

processing the first detection output to determine one or more modified detection characteristics of said plurality of received detection characteristics, by:

retrieving a mapping function; and

applying the mapping function, the mapping being dependent upon at least one of the plurality of received detection characteristics.

2. A method according to claim 1 , wherein the plurality of received detection characteristics comprises at least one of:

a detection score associated with a detection of an object in the image;

an estimated location of an object relative to the first region of the image; and

estimated dimensions of an object in the image.

3. A method according to claim 2 , wherein the one or more modified detection characteristics comprise a modified estimated detection score associated with the detection of the object in the image.

4. A method according to claim 1 , wherein the processing of the first detection output to determine the one or more modified detection characteristics comprises processing one or more further detection outputs, each of the further detection outputs being associated with a respective further region of the image.

5. A method according to claim 4 , wherein a further region of the image associated with one of the further detection outputs is adjacent to the first region of the image.

6. A method according to claim 4 , wherein a further region of the image associated with one of the further detection outputs has different dimensions to the first region of the image.

7. A method according to claim 1 , wherein the plurality of received detection outputs comprises:

an estimated location of an object relative to the first region of the image; and

estimated dimensions of an object in the image,

wherein the processing of the first detection output comprises:

determining, from the estimated location of the object and the estimated dimensions of the object, that a portion of the object extends outside of a border of the image; and

rejecting the detection of the object on the basis of the determination of the portion of the object extending outside of the border of the image.

8. A method according to claim 1 , wherein the plurality of received detection outputs comprises:

a detection score associated with a detection of an object in the image;

an estimated location of an object relative to the first region of the image; and

estimated dimensions of an object in the image,

wherein the processing of the first detection output comprises:

determining, from the estimated location of the object and the estimated dimensions of the object, that a portion of the object extends outside of a border of the image; and

modifying the detection score on the basis of the determination of the portion of the object extending outside of the border of the image.

9. A method according to claim 1 , wherein:

the object detection system generates the object detection data by processing image data received from an associated image capture system, and

the mapping is independent of the associated image capture system.

10. A method according to claim 9 , comprising:

processing the one or more modified detection characteristics to determine at least one corrected modified detection characteristic, by:

retrieving a further mapping function; and

applying the further mapping function, the further mapping being dependent upon at least one of the one or more modified detection characteristics.

11. A method of calibrating a system for modifying data generated by an object detection system, the method comprising:

receiving a calibration image comprising a first object with predetermined object characteristics;

processing the calibration image, using the object detection system, to generate a plurality of detection outputs associated with different respective regions of the calibration image and each comprising a respective plurality of detection characteristics;

associating the plurality of detection outputs with the predetermined object characteristics to generate a labeled plurality of detection outputs; and

processing the labeled plurality of detection outputs using supervised learning to generate a mapping function,

wherein the mapping function is configured to process one or more received detection characteristics to determine one or more modified detection characteristics, the mapping being dependent upon at least one of the one or more received detection characteristics.

12. A method according to claim 11 , wherein the predetermined object characteristics comprise at least one of:

a location of the first object within the calibration image; and

dimensions of the first object.

13. A method according to claim 11 , wherein the processing of the calibration image to generate the plurality of detection outputs comprises processing the calibration image with a model of an image capture system associated with the object detection system.

14. A non-transient storage medium comprising machine readable instructions which, when executed by processing circuitry of a computer system, cause the computer system to perform a method of modifying data comprising a plurality of detection outputs associated with different respective regions of an image, wherein a first detection output of the plurality of detection outputs is associated with a first region of an image and comprises a plurality of received detection characteristics,

the method comprising processing the first detection output to determine one or more modified detection characteristics of said plurality of detection characteristics, by:

retrieving a mapping function; and

applying the mapping function, the mapping being dependent upon at least one of the plurality of received detection characteristics.