Systems and methods for facilitating navigation in space
A navigation system for facilitating navigation in a space includes a network interface configured to access a representation of a space, a memory storing instructions, and at least one processor coupled with the memory. The at least one processor executes the instructions to detect objects in the representation of the space and determine a plurality of desirable positions which facilitate navigation in the representation of the space based on the detected objects.
1. A navigation system for facilitating navigation in a space, the navigation system comprising:
a network interface configured to access a representation of a space;
a memory storing instructions; and
at least one processor coupled with the memory and executing the instructions to:
detect objects in the representation of the space;
determine a plurality of navigation positions which facilitate navigation in the representation of the space based on the detected objects;
determine a plurality of navigational positions in the representation of the space;
render a panoramic image from at least one of the plurality of navigational positions; and
calculate a scene perception (SP) score at the at least one of the plurality of navigational positions based on a generated panoramic image,
wherein the plurality of navigation positions is determined by:
selecting a group of navigation positions, of which the SP scores are greater than a predetermined threshold, from among the plurality of navigational positions; and
generating a teleport graph by connecting the group of navigation positions with edges.
2. The navigation system according to claim 1 , wherein the space is a virtual, augmented, mixed reality or real space.
3. The navigation system according to claim 1 , wherein the plurality of navigational positions is determined in a free space, which is free or clear of the detected objects in the representation of the space.
4. The navigation system according to claim 1 , wherein the plurality of navigational positions is determined based on a unit area, wherein the unit area is based on a free space, which is free or clear of the objects in the representation of the space, and a predetermined neighboring area from at least one of the plurality of navigational positions.
5. The navigation system according to claim 1 , wherein the SP score at the at least one of the plurality of navigational positions is based on visually desirable views.
6. The navigation system according to claim 1 , wherein the panoramic image is an observer's first person view at the at least one of the plurality of navigational positions.
7. The navigation system according to claim 1 , wherein the teleport graph includes the group of navigation positions and edges connecting at least one navigation position of the group of navigation positions to another desirable position of the group of navigation positions.
8. The navigation system according to claim 7 , wherein the plurality of navigation positions of the teleport graph are connected via edges.
9. The navigation system according to claim 7 , wherein the at least one processor executes the instructions further to:
refine the teleport graph based on cost functions;
repeat generation of the group of navigation positions, connecting the group of the navigation positions, and refinement of the teleport graph until a change in a total cost of the teleport graph converges; and
output a final version teleport graph after convergence of the total cost.
10. The navigation system according to claim 9 , wherein refinement is performed by the at least one processor by:
adding a position, which is in the plurality of navigational positions but not in the teleport graph, to the teleport graph;
deleting a navigation position of the plurality of navigation positions of the teleport graph; or
moving a position of one navigation position of the plurality of navigation positions of the teleport graph to a position within a neighborhood of the plurality of navigation positions.
11. The navigation system according to claim 1 , wherein the at least one processor executes the instructions further to:
generate a reference panoramic view based on:
determining a metric based on evaluating how similar a new view at a given position is to a plurality of human marked positions;
providing the metric as an input into a machine learning model that predicts similarities;
assigning a scene perception score based on the predicted similarities; and
generating the reference panoramic view based on the assigned scene perception score.
12. The navigation system according to claim 1 , wherein the scene perception (SP) score is calculated by a machine learning network.
13. A method for facilitating navigation in a space, the method comprising
accessing a representation of a space;
detecting objects in the representation of the space;
determining a plurality of navigation positions, which facilitate navigation in the representation of the space based on the detected objects;
determining a plurality of navigational positions in the representation of the space;
generating a panoramic image from at least one of the plurality of navigational positions;
calculating a scene perception (SP) score at the at least one of the plurality of navigational positions based on a generated panoramic image; and
generating a reference panoramic view based on:
determining a metric based on evaluating how similar a new view at a given position is to a plurality of human marked positions;
providing the metric as an input into a machine learning model that predicts similarities;
assigning a scene perception score based on the predicted similarities; and
generating the reference panoramic view based on the assigned scene perception score.
14. The method according to claim 13 , wherein the plurality of navigation positions is determined by
selecting a group of navigation positions, of which the SP scores are greater than a predetermined threshold, from among the plurality of navigational positions; and
generating a teleport graph by connecting the group of navigation positions with edges.
15. The method according to claim 14 , further comprising:
refining the teleport graph based on cost functions;
repeating generation of the group of navigation positions, connecting the group of the navigation positions, and refinement of the teleport graph until a change in a total cost of the teleport graph converges; and
outputting a final version teleport graph after convergence of the total cost.
16. The method according to claim 15 , wherein refinement is performed by:
adding a position, which is in the plurality of navigational positions but not in the teleport graph, to the teleport graph;
deleting a navigation position in the teleport graph; or
moving a position of one navigation position in the teleport graph to a position within a neighborhood of the position.
17. A non-transitory computer-readable medium storing instructions that, when executed by a computer, cause the computer to perform a method for facilitating navigation in a space, the method comprising:
accessing a representation of a space;
detecting objects in the representation of the space;
determining a plurality of navigation positions, which facilitate navigation in the representation of the space based on the detected objects;
determining a plurality of navigational positions in the representation of the space;
generating a panoramic image from at least one of the plurality of navigational positions;
calculating a scene perception (SP) score at the at least one of the plurality of navigational positions based on a generated panoramic image; and
generating a reference panoramic view based on:
determining a metric based on evaluating how similar a new view at a given position is to a plurality of human marked positions;
providing the metric as an input into a machine learning model that predicts similarities;
assigning a scene perception score based on the predicted similarities; and
generating the reference panoramic view based on the assigned scene perception score.