Optical Neuromorphic Eikonal Solver - Benchmark Datasets

# Optical Neuromorphic Eikonal Solver - Benchmark Datasets

Benchmark datasets for evaluating GPU-accelerated pathfinding algorithms.

## Content

5 synthetic pathfinding test cases:
- sparse_128.npz (128×128, 10% obstacles)
- medium_256.npz (256×256, 20% obstacles)
- gradient_256.npz (256×256, gradient speeds)
- maze_511.npz (511×511, perfect maze)
- complex_512.npz (512×512, 30% obstacles)

Plus benchmark results showing GPU solver performance.

## Performance

- 134.9× average speedup vs CPU Dijkstra
- 0.64% mean absolute error
- 1.025× optimal path length
- 2-4ms per query on 512×512 grids

## Format

NumPy .npz archives containing:
- obstacles: (H,W) float32 array
- speeds: (H,W) float32 array
- source: (2,) int32 array
- target: (2,) int32 array
- metadata: JSON string

See DATASETS.md for complete specification.

## Links

- Code: https://github.com/Agnuxo1/Optical-Neuromorphic-Computing-for-Real-Time-Pathfinding-A-GPU-Accelerated-Eikonal-Solver
- Paper: https://github.com/Agnuxo1/Optical-Neuromorphic-Computing-for-Real-Time-Pathfinding-A-GPU-Accelerated-Eikonal-Solver/blob/main/optical_neuromorphic_paper.html
- Datasets: https://huggingface.co/datasets/Agnuxo/optical-neuromorphic-eikonal-benchmarks
- Demo: https://huggingface.co/spaces/Agnuxo/optical-neuromorphic-pathfinding-demo

## Citation

See README or paper for citation information.