Solid-state LiDAR public benchmark v0.3¶
This report summarizes the checked public-data-only run. It is a ground-truth-free temporal-holdout comparison, not an absolute extrinsic or clock-accuracy result.
- Schema:
slac.solid_state_cross_dataset_benchmark/v0.2 - Tool:
tools/run_solid_state_cross_dataset_benchmark.py(0.4.0) - Spec:
examples/public_datasets/solid_state_cross_dataset_benchmark_v03.yaml - Spec SHA-256:
4eca1e7668ddca4fdcb206daaf9d5e99870779ec0292dcc41d2ce64506531267 - Provenance sources:
142
Summary¶
- Scored datasets: 3/3
- Scored replicates: 27/27
- Adaptive wins: 23/27 (
0.852) - Mean holdout improvement: 47.70%
- Median holdout improvement: 65.94%
- Bootstrap 95% CI: [34.43, 59.99]%
| Dataset | Replicates | Adaptive win rate | Mean improvement (95% CI) | Interpretation |
|---|---|---|---|---|
| AgRob Modular-e Livox MID-70 ↔ RS-LiDAR | 9 | 0.556 | −1.34% (−10.16, 6.68) | mixed; retained counterexample |
| TIERS LidarsCali VLP-16 ↔ Livox | 9 | 1.000 | +78.76% (+76.00, 81.73) | adaptive wins all |
| AIST GLIM identity control | 9 | 1.000 | +65.67% (+64.18, 66.95) | adaptive wins all |
AgRob is not hidden or treated as a failure of the dataset: its result is simply mixed across temporal boundaries and deterministic seeds. The result supports the declared paired temporal-holdout protocol, not a universal SOTA claim. The public fixtures do not provide independently surveyed extrinsic or absolute clock ground truth.
Reproduction¶
Run the four-replicate v0.2 pilot first, then extend it with the three-way, three-seed v0.3 declaration:
python tools/run_solid_state_benchmark_replicates.py `
examples/public_datasets/solid_state_cross_dataset_benchmark.yaml `
--output-root outputs/solid_state_benchmark_v02_replicates `
--output-spec examples/public_datasets/solid_state_cross_dataset_benchmark_v02.yaml `
--split-id middle_holdout --split-id late_holdout `
--seed 0 --seed 17
python tools/run_solid_state_benchmark_replicates.py `
examples/public_datasets/solid_state_cross_dataset_benchmark.yaml `
--output-root outputs/solid_state_benchmark_v03_replicates `
--output-spec examples/public_datasets/solid_state_cross_dataset_benchmark_v03.yaml `
--append-spec examples/public_datasets/solid_state_cross_dataset_benchmark_v02.yaml `
--split-id early_holdout --split-id middle_holdout --split-id late_holdout `
--seed 0 --seed 17 --seed 42
python tools/run_solid_state_cross_dataset_benchmark.py `
examples/public_datasets/solid_state_cross_dataset_benchmark_v03.yaml `
--output outputs/solid_state_cross_dataset_benchmark_v03.yaml `
--markdown-output outputs/solid_state_cross_dataset_benchmark_v03.md `
--html-output outputs/solid_state_cross_dataset_benchmark_v03.html