Neural networks for object detection

A computing system, comprising:

• one or more processing devices;
• and
• one or more computer-readable media encoded with instructions that, when executed by the one or more processing devices, cause the one or more processing devices to perform operations comprising:
• for each time step after an initial time step in a series of time steps:
• obtaining a first input that represents an input image, wherein the input image depicts a set of objects at various positions of the input image;
• obtaining a second input that identifies respective positions in the input image of a subset of objects from the set of objects;
• determining, by processing the first input and the second input, with a neural network, a position in the input image of an additional object that is not among the subset of objects;
• and
• adding the position of the additional object to the second input so that the additional object is among the subset of objects in the next time step.

One or more non-transitory computer-readable media encoded with instructions that, when executed by one or more processing devices, cause the one or more processing devices to perform operations comprising:

• for each time step after an initial time step in a series of time steps:
• obtaining a first input that represents an input image, wherein the input image depicts a set of objects at various positions of the input image;
• obtaining a second input that identifies respective positions in the input image of a subset of objects from the set of objects;
• determining, by processing the first input and the second input, with a neural network, a position in the input image of an additional object that is not among the subset of objects;
• and
• adding the position of the additional object to the second input so that the additional object is among the subset of objects in the next time step.