CLI Reference

VigilCV ships with a powerful terminal interface powered by Typer and Rich.

pip install vigilcv
vigilcv --help

---

vigilcv inspect

Inspect a single image and print a detailed quality report.

vigilcv inspect IMAGE_PATH [OPTIONS]

Arguments: Options: Example:

vigilcv inspect photo.jpg --blur-threshold 150 --json

Output:

{
  "path": "photo.jpg",
  "blur_score": 423.7,
  "shannon_entropy": 6.12,
  "underexposure_ratio": 0.03,
  "overexposure_ratio": 0.01,
  "passed": true,
  "warnings": []
}

---

vigilcv audit

Audit all images in a directory with parallel workers.

vigilcv audit DIRECTORY [OPTIONS]

Options: Example:

vigilcv audit dataset/train/ --workers 16 --report report.html

Terminal Output:

 VigilCV BatchAuditor
 ──────────────────────────────────────────
  Total images   : 12,847
  Valid          : 12,391  (96.5%)
  Corrupted      : 23
  Blurred        : 341
  Underexposed   : 82
  Overexposed    : 10
  Throughput     : 3,241 FPS
 ──────────────────────────────────────────
  HTML report saved: report.html

---

vigilcv baseline

Compute and save a drift baseline from a reference dataset.

vigilcv baseline DIRECTORY --output BASELINE_PATH

Options: Example:

vigilcv baseline dataset/train/ --output baselines/production_v1.pkl

---

vigilcv drift

Detect distribution drift against a saved baseline.

vigilcv drift DIRECTORY --baseline BASELINE_PATH [OPTIONS]

Options: Example:

vigilcv drift dataset/new_batch/ --baseline baselines/production_v1.pkl --threshold 0.20

Output:

 Drift Report
 ──────────────────────────────────────────
  Status         : DRIFTED
  Wasserstein-1  : 0.287
  MMD²           : 0.00412
  Reference N    : 12,847
  Query N        : 1,200
  Top features   : channel_r_mean, spatial_q1_g_std, kurtosis_b