Neural network-based maps

1 . A processor comprising:

one or more circuits to use one or more updated neural networks to generate three-dimensional (3D) information based, at least in part, on image data, wherein to generate the 3D information, the one or more circuits:

generate a plurality of viewpoints that include at least one map feature predicted by the one or more updated neural networks that was not predicted by a prior version of the one or more updated neural networks, wherein the at least one map feature is identified by a second one or more neural networks using prior viewpoints generated by the prior version of the one or more updated neural networks, and wherein one or more neural fields of the one or more updated neural networks encode the at least one map feature; and

based, at least in part, on the plurality of viewpoints, merge information of the plurality of viewpoints corresponding to the at least one map feature to generate the 3D information.

2 . The processor of claim 1 , wherein:

the second one or more neural networks are to be used to identify one or more objects using one or more neural radiance fields utilizing the prior viewpoints generated by the prior version of the one or more updated neural networks, wherein the at least one map feature is associated with at least an object of the identified one or more objects; and

the one or more updated neural networks are to encode the 3D information based, at least in part, on the identified one or more objects using the second one or more neural networks.

3 . The processor of claim 1 , wherein the prior version of the one or more updated neural networks are to be updated based, at least in part, on one or more images of an area collected during one or more trips through the area.

4 . The processor of claim 1 , wherein a decoder is to generate one or more vectors of map information based, at least in part, on the one or more updated neural networks.

5 . The processor of claim 1 , wherein:

one or more neural radiance fields are to encode a 3D representation of an area to be mapped; and

wherein updating the prior version of the one or more updated neural networks causes the one or more updated neural networks to generate the 3D information using the one or more neural radiance fields.

6 . The processor of claim 1 , wherein the one or more updated neural networks are to generate the 3D information based, at least in part, on alignment of the 3D information encoded by two or more third neural networks.

7 . A system comprising:

one or more processors to use one or more updated neural networks to generate three-dimensional (3D) information based, at least in part, on image data, wherein to generate the 3D information, the one or more processors:

generate a plurality of viewpoints that include at least one map feature predicted by the one or more updated neural networks that was not predicted by a prior version of the one or more updated neural networks, wherein the at least one map feature is identified by a second one or more neural networks using prior viewpoints generated by the prior version of the one or more updated neural networks, and wherein one or more neural fields of the one or more updated neural networks encode the at least one map feature; and

based, at least in part, on the plurality of viewpoints, merge information of the plurality of viewpoints corresponding to the at least one map feature to generate the 3D information.

8 . The system of claim 7 , wherein:

the one or more processors are to use the second one or more neural networks to identify one or more objects using one or more neural radiance fields utilizing the prior viewpoints generated by the prior version of the one or more updated neural networks, wherein the at least one map feature is associated with at least an object of the identified one or more objects; and

the one or more processors are to use the one or more updated neural networks to encode the 3D information based, at least in part, on the identified one or more objects using the second one or more neural networks.

9 . The system of claim 7 , wherein the prior version of the one or more updated neural networks are to be updated based, at least in part, on one or more images of an area collected during one or more trips through the area.

10 . The system of claim 7 , wherein a decoder is to generate one or more vectors of map information based, at least in part, on the one or more updated neural networks.

11 . The system of claim 7 , wherein:

one or more neural radiance fields are to encode a 3D representation of an area to be mapped; and

wherein updating of the prior version of the one or more updated neural networks causes the one or more updated neural networks to generate the 3D information using the one or more neural radiance fields.

12 . The system of claim 7 , wherein the one or more updated neural networks are to generate the 3D information based, at least in part, on alignment of the 3D information encoded by two or more third neural networks.

13 . A method comprising:

using one or more updated neural networks to generate three-dimensional (3D) information based, at least in part, on image data, wherein generation of the 3D information includes:

generating a plurality of viewpoints that include at least one map feature predicted by the one or more updated neural networks that was not predicted by a prior version of the one or more updated neural networks, wherein the at least one map feature is identified by a second one or more neural networks using prior viewpoints generated by the prior versions of the one or more updated neural networks, and wherein one or more neural fields of the one or more updated neural networks encode the at least one map feature; and

based, at least in part, on the plurality of viewpoints, merging information of the plurality of viewpoints corresponding to the at least one map feature to generate the 3D information.

14 . The method of claim 13 , further comprising:

using the one or more second neural networks to identify one or more objects encoded by one or more neural radiance fields utilizing the prior viewpoints generated by the prior version of the one or more updated neural networks, wherein the at least one map feature is associated with at least an object of the identified one or more objects; and

causing the one or more updated neural networks to encode the 3D information based, at least in part, on the identified one or more objects using the one or more second neural networks.

15 . The method of claim 13 , further comprising:

updating the prior version of the one or more updated neural networks based, at least in part, on one or more images of an area collected during one or more trips through the area.

16 . The method of claim 13 , further comprising:

using a decoder to generate one or more vectors of map information based, at least in part, on the one or more updated neural networks.

17 . The method of claim 13 , further comprising:

using one or more neural radiance fields to encode a 3D representation of an area to be mapped; and

wherein updating of the prior version of the one or more neural updated networks causes the one or more updated neural networks to generate the 3D information using the one or more neural radiance fields.