From 3cd9138c9d8fb9be248c75330cca9da47f4d2e8a Mon Sep 17 00:00:00 2001
From: Lukasz Kasprzak <lukas@labunix.xyz>
Date: Sun, 10 May 2026 22:57:02 +0200
Subject: added a whole bunch of scripts

---
 ocr-scan | 146 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
 1 file changed, 146 insertions(+)
 create mode 100755 ocr-scan

(limited to 'ocr-scan')

diff --git a/ocr-scan b/ocr-scan
new file mode 100755
index 0000000..36994b5
--- /dev/null
+++ b/ocr-scan
@@ -0,0 +1,146 @@
+#!/usr/bin/env bash
+# ocr-scan.sh — recursively OCR un-OCR'd PDFs
+# Deps: ocrmypdf, poppler-utils (pdftotext), tesseract-ocr
+# Optional: python3-langdetect (for -a auto-detect)
+
+set -euo pipefail
+
+LANG_CODE="eng"
+AUTO_DETECT=false
+MIN_CHARS=50        # chars per page threshold — below = needs OCR
+DIR="."
+
+usage() {
+    cat <<EOF
+Usage: $(basename "$0") [OPTIONS] [DIRECTORY]
+
+Recursively find un-OCR'd PDFs and produce OCR'd copies with _OCR suffix.
+
+OPTIONS:
+  -l LANG     Tesseract language code(s)  (default: eng)
+              Single:       -l pol
+              Multiple:     -l pol+eng
+              Examples:     eng pol tha deu fra rus chi_sim jpn ara
+  -a          Auto-detect language via langdetect (requires: pip install langdetect)
+              Note: falls back to eng for fully un-OCR'd scans (no text layer)
+  -h          Show this help and exit
+
+ARGUMENTS:
+  DIRECTORY   Root directory to search recursively (default: current directory)
+
+OUTPUT:
+  Each processed file is saved alongside the original:
+    original.pdf → original_OCR.pdf
+  Files already having an _OCR copy or sufficient text are skipped.
+
+DEPENDENCIES:
+  Required:   ocrmypdf, poppler-utils (pdftotext, pdfinfo), tesseract-ocr
+  Per-lang:   apt install tesseract-ocr-pol tesseract-ocr-tha ...
+  Optional:   pip install langdetect  (for -a flag)
+
+EXAMPLES:
+  $(basename "$0")                        # current dir, English
+  $(basename "$0") ~/docs                 # ~/docs recursively, English
+  $(basename "$0") -l pol ~/docs          # Polish
+  $(basename "$0") -l pol+eng ~/docs      # Polish + English mixed
+  $(basename "$0") -a ~/docs             # auto-detect language
+EOF
+    exit 0
+}
+
+while getopts ":l:ah" opt; do
+    case $opt in
+        l) LANG_CODE="$OPTARG" ;;
+        a) AUTO_DETECT=true ;;
+        h) usage ;;
+        :) echo "Option -$OPTARG requires an argument."; usage ;;
+        \?) echo "Unknown option: -$OPTARG"; usage ;;
+    esac
+done
+shift $((OPTIND - 1))
+[[ $# -gt 0 ]] && DIR="$1"
+
+# Map ISO 639-1 → Tesseract lang codes
+map_lang() {
+    case "$1" in
+        pl|pol) echo "pol" ;;
+        en|eng) echo "eng" ;;
+        th|tha) echo "tha" ;;
+        de|deu) echo "deu" ;;
+        fr|fra) echo "fra" ;;
+        ru|rus) echo "rus" ;;
+        zh|chi) echo "chi_sim" ;;
+        ja|jpn) echo "jpn" ;;
+        ar|ara) echo "ara" ;;
+        *)       echo "eng" ;;
+    esac
+}
+
+detect_lang() {
+    local pdf="$1"
+    local sample
+    sample=$(pdftotext "$pdf" - 2>/dev/null | head -c 500)
+    if [[ -z "$sample" ]]; then
+        echo "eng"; return
+    fi
+    python3 - <<EOF 2>/dev/null || echo "eng"
+from langdetect import detect
+try:
+    print("$(echo "$sample" | tr -d "'" | head -c 300)")
+except:
+    print("eng")
+EOF
+    # simpler: call python directly
+    python3 -c "
+from langdetect import detect
+import sys
+try:
+    lang = detect('''$sample''')
+    print(lang)
+except:
+    print('eng')
+" 2>/dev/null || echo "eng"
+}
+
+needs_ocr() {
+    local pdf="$1"
+    local pages
+    pages=$(pdfinfo "$pdf" 2>/dev/null | awk '/^Pages:/{print $2}')
+    [[ -z "$pages" || "$pages" -eq 0 ]] && pages=1
+    local chars
+    chars=$(pdftotext "$pdf" - 2>/dev/null | wc -c)
+    local threshold=$(( pages * MIN_CHARS ))
+    [[ "$chars" -lt "$threshold" ]]
+}
+
+find "$DIR" -type f -iname "*.pdf" ! -iname "*_OCR.pdf" | while read -r pdf; do
+    out="${pdf%.pdf}_OCR.pdf"
+
+    if [[ -f "$out" ]]; then
+        echo "[SKIP] Already has OCR copy: $out"
+        continue
+    fi
+
+    if ! needs_ocr "$pdf"; then
+        echo "[SKIP] Already has text: $pdf"
+        continue
+    fi
+
+    lang="$LANG_CODE"
+    if $AUTO_DETECT; then
+        detected=$(detect_lang "$pdf")
+        lang=$(map_lang "$detected")
+        echo "[DETECT] $pdf → $detected → $lang"
+    fi
+
+    echo "[OCR] $pdf (lang: $lang) → $out"
+    ocrmypdf \
+        --language "$lang" \
+        --output-type pdfa \
+        --skip-text \
+        --rotate-pages \
+        --deskew \
+        "$pdf" "$out" \
+        && echo "[DONE] $out" \
+        || echo "[FAIL] $pdf"
+done
-- 
cgit v1.3