Scripts Reference

All executable scripts live in scripts/ and are run from the repository root.

train.py -- STDP Training

Layer-wise STDP training with WTA competition and adaptive thresholds.

python scripts/train.py --config configs/config.yaml [OPTIONS]

Flag	Default	Description
--config, -c	--	YAML config file
--paper	--	Paper preset (igarss2023, kheradpisheh2018)
--output, -o	./runs	Output directory
--name, -n	auto-generated	Experiment name
--resume, -r	--	Resume from checkpoint
--epochs	50	Maximum epochs per layer
--n-classes	1	Number of output classes
--device	cuda	cuda, cpu, cuda:0
--seed	42	Random seed
--log-level	INFO	Logging verbosity
--no-tensorboard	false	Disable TensorBoard

Outputs: checkpoints (checkpoint_best.pt, checkpoint_latest.pt, per-epoch), metrics_final.json, TensorBoard logs.

---

train_cv.py -- Cross-Validation

k-fold cross-validation with held-out test set.

python scripts/train_cv.py --config configs/config.yaml --n-folds 10 [OPTIONS]

Flag	Default	Description
--config, -c	configs/config.yaml	Config file
--data-root, -d	from config	EBSSA data root
--n-folds, -k	10	Number of folds
--test-ratio	0.1	Held-out test fraction
--seed	42	Split seed
--output-dir, -o	runs/cv	Output directory
--threshold	0.05	Inference threshold
--eval-only	false	Skip training, evaluate existing folds

Outputs: fold_info.json, test_recordings.txt, per-fold directories, cv_results.json.

---

evaluate.py -- Model Evaluation

Pixel-level, object-level, or volume-based evaluation.

python scripts/evaluate.py --checkpoint best.pt --data-root /path/to/EBSSA [OPTIONS]

Flag	Default	Description
--checkpoint, -c	required	Checkpoint path
--data-root, -d	--	EBSSA root
--split	val	train, val, test, all
--sensor	all	ATIS, DAVIS, all
--volume-based	false	IGARSS 2023 methodology
--object-level	false	Centroid matching
--spatial-tolerance	1	TP tolerance in pixels
--no-hulk	false	Disable HULK
--no-smash	false	Disable SMASH
--output, -o	--	Save metrics JSON
--debug	false	Debug logging

---

inference.py -- Detection and Segmentation

python scripts/inference.py --checkpoint best.pt --input recording.mat [OPTIONS]

Flag	Default	Description
--checkpoint, -c	required	Checkpoint path
--input, -i	--	Single .mat file
--data-root, -d	--	Batch mode: EBSSA root
--output, -o	detections.json	Output JSON
--threshold	0.05	Inference threshold
--visualize	false	Save 2D images
--visualize-3d	false	3D trajectory plots
--device	auto	cuda or cpu

Outputs: detections.json, optional detection_*.png and figure4_sample_*.png.

---

threshold_sweep.py -- Threshold Optimization

Sweeps inference thresholds to find optimal informedness.

python scripts/threshold_sweep.py --checkpoint best.pt --data-root /path/to/EBSSA

Tests thresholds: [0.1, 0.2, 0.3, 0.5, 0.7, 1.0, 1.5, 2.0, 3.0, 5.0]. Prints a table with sensitivity, specificity, and informedness for each, plus the best threshold.

---

diagnose.py -- Diagnostic Visualization

python scripts/diagnose.py --checkpoint best.pt --data-root /path/to/EBSSA --n-samples 5

Flag	Default	Description
--checkpoint	hard-coded path	Checkpoint
--data-root	hard-coded path	EBSSA root
--n-samples	5	Number of samples

Output: diagnostic_output.png -- 5-column grid per sample (input events, GT, raw spikes, overlay, classification result).

---

demo.py -- Quick Demonstration

python scripts/demo.py

Placeholder script for running a quick forward-pass demonstration. Currently empty; implement to showcase the full pipeline on a sample recording.