Deep geometric model fitting
Systems, apparatuses and methods may provide for technology that generates, by a first neural network, an initial set of model weights based on input data and iteratively generates, by a second neural network, an updated set of model weights based on residual data associated with the initial set of model weights and the input data. Additionally, the technology may output a geometric model of the input data based on the updated set of model weights. In one example, the first neural network and the second neural network reduce the dependence of the geometric model on the number of data points in the input data.
1. A system comprising:
a memory to store input data associated with a plurality of cameras; and
a semiconductor apparatus coupled to the memory, the semiconductor apparatus including one or more substrates and logic coupled to the one or more substrates, the logic coupled to the one or more substrates to:
generate, by a first neural network, an initial group of model weights based on the input data,
execute, by a second neural network, a repeated process to generate an updated group of model weights based on residual data associated with the initial group of model weights and the input data, and
output a geometric model of the input data based on the updated group of model weights.
2. The system of claim 1 , further comprising:
a processor to localize the plurality of cameras in a three-dimensional environment based on the geometric model.
3. The system of claim 1 , wherein the geometric model is associated with a vehicular functionality.
4. The system of claim 1 , wherein the geometric model is associated with a robotic functionality.
5. The system of claim 1 , wherein the geometric model is associated with a motion pattern.
6. The system of claim 1 , wherein the repeated process is an iterative process.
7. A semiconductor apparatus comprising:
one or more substrates; and
logic coupled to the one or more substrates, wherein the logic is implemented at least partly in one or more of configurable logic or fixed-functionality hardware logic, the logic coupled to the one or more substrates to:
generate, by a first neural network, an initial group of model weights based on input data, wherein the input data is associated with a plurality of cameras;
execute, by a second neural network, a repeated process to generate an updated group of model weights based on residual data associated with the initial group of model weights and the input data; and
output a geometric model of the input data based on the updated group of model weights.
8. The apparatus of claim 7 , wherein the logic coupled to the one or more substrates is to localize the plurality of cameras in a three-dimensional environment based on the geometric model.
9. The apparatus of claim 7 , wherein the geometric model is associated with a vehicular functionality.
10. The apparatus of claim 7 , wherein the geometric model is associated with a robotic functionality.
11. The apparatus of claim 7 , wherein the geometric model is associated with a motion pattern.
12. The apparatus of claim 7 , wherein the repeated process is an iterative process.
13. The apparatus of claim 7 , wherein the logic coupled to the one or more substrates includes transistor channel regions that are positioned within the one or more substrates.
14. At least one non-transitory computer readable storage medium comprising a group of instructions, which when executed by a computing device, cause the computing device to:
generate, by a first neural network, an initial group of model weights based on input data, wherein the input data is associated with a plurality of cameras;
execute, by a second neural network, a repeated process to generate an updated group of model weights based on residual data associated with the initial group of model weights and the input data; and
output a geometric model of the input data based on the updated group of model weights.
15. The at least one non-transitory computer readable storage medium of claim 14 , wherein the instructions, when executed, cause the computing device to localize the plurality of cameras in a three-dimensional environment based on the geometric model.
16. The at least one non-transitory computer readable storage medium of claim 14 , wherein the geometric model is associated with a vehicular functionality.
17. The at least one non-transitory computer readable storage medium of claim 14 , wherein the geometric model is associated with a robotic functionality.
18. The at least one non-transitory computer readable storage medium of claim 14 , wherein the geometric model is associated with a motion pattern.
19. The at least one non-transitory computer readable storage medium of claim 14 , wherein the repeated process is an iterative process.
20. A method comprising:
generating, by a first neural network, an initial group of model weights based on input data, wherein the input data is associated with a plurality of cameras;
executing, by a second neural network, a repeated process to generate an updated group of model weights based on residual data associated with the initial group of model weights and the input data; and
outputting a geometric model of the input data based on the updated group of model weights.
21. The method of claim 20 , further comprising localizing the plurality of cameras in a three-dimensional environment based on the geometric model.
22. The method of claim 20 , wherein the geometric model is associated with a vehicular functionality.
23. The method of claim 20 , wherein the geometric model is associated with a robotic functionality.
24. The method of claim 20 , wherein the geometric model is associated with a motion pattern.
25. The method of claim 20 , wherein the repeated process is an iterative process.