initial

Makefile

@@ -0,0 +1,12 @@
+PREFIX = /usr/local
+
+all: install
+
+install:
+	install -D -m 0755 gotcha ${DESTDIR}${PREFIX}/bin/gotcha
+
+uninstall:
+	rm ${DESTDIR}${PREFIX}/bin/gotcha
+
+test:
+	sh ./test.sh

README.md

@@ -0,0 +1,3 @@
+# gotcha
+
+playing around with captcha breaking

gotcha_v1

@@ -0,0 +1,143 @@
+#!/bin/sh -e
+#
+# Breaks simple captchas.
+#
+# deps: gocr, imagemagick, multicrop, textcleaner
+#
+
+# -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
+NUM_CHARS=4
+CHARSET="[0-9A-z]"
+
+# lower number = more dense to search = more rigorous detection
+DENSITY=3
+
+# 0-100, higher numbers will force strict matches
+CERTAINTY=0
+
+# 16: not dot divide overlapping chars
+# 32: do not context correct
+MODE='-m 16 -m 32'
+
+IMAGE_PROG=mpvimg
+# -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
+
+usage() {
+    >&2 printf "Usage: %s [-d] [-v] [file]\n" "${0##*/}"
+    exit 1
+}
+
+init() {
+    random() { r=$(($(date +%s) / $$ - $$)) ; echo ${r#-} ; }
+    dir=/tmp/$(random) # dir to store tmp files
+    tmp=$dir/$(random) # image created
+    unset -f random
+
+    clean
+    mkdir -p "$dir"
+}
+
+clean() {
+    rm -r "${dir:-?}" 2>/dev/null ||:
+}
+
+check() {
+    if [ ! -f "$1" ] ; then
+        >&2 echo "No such image: $1"
+        exit 1
+    fi
+}
+
+jpg2png() {
+    case $(file "$1") in
+        *jpg*|*JPG*|*jpeg*|*JPEG*)
+            convert "$1" "${1%.*}.png"
+            rm "$1"
+            image=${image%.*}.png
+            ;;
+    esac
+}
+
+main() {
+    image=$1
+    check "$image"
+
+    init
+    # trap 'clean' INT TERM EXIT
+
+    # convert to png if necessary
+    jpg2png "image"
+
+    # clean up text and $IM<Plug> black on white background
+    # text will be evenly spaced but not aligned
+    convert \
+        -quality 100 \
+        -shave 1x1 \
+        -resize 500%x500% \
+        -threshold 50% \
+        -deskew 50 \
+        -trim \
+        -bordercolor white \
+        -border 20 \
+        "$image" "$tmp"
+
+    [ "$DEBUG" = true ] && $IMAGE_PROG "$tmp" &
+
+
+    # IMPORTANT: multicrop reads from top-to-bottom, while
+    # we need to read left-to-right as we are reading text.
+    #
+    # Without rotation the resulting letters are mixed at random.
+    convert -rotate 90 "$tmp" "$tmp"
+    multicrop -u 1 -f 20 -g "${DENSITY:-5}" \
+        -b white "$tmp" "$dir/multicut.png" >/dev/null
+
+    # expand borders a little to help with ocr
+    for i in "$dir"/multicut-*.png ; do
+        {
+            # and rotate 270 to put us back where we were
+            convert -rotate 270 "$i" "$i"
+
+            textcleaner -g -e stretch -f 25 -o 10 -s 1 "$i" "$i"
+
+            convert \
+                -quality 100 \
+                -monochrome \
+                -trim \
+                -bordercolor white \
+                -border 100 \
+                "$i" "$i"
+        } &
+    done
+    wait
+
+    if [ "$DEBUG" = true ] ; then
+        for i in "$dir/multicut-"*.png ; do
+            $IMAGE_PROG "$i" &
+        done
+    fi
+
+    montage "$dir/multicut-*.png" -tile "${NUM_CHARS:-4}"x1 "$dir/out.png"
+
+    if [ "$DEBUG" = true ] || [ "$DISPLAY_RESULT" = true ] ; then
+        $IMAGE_PROG "$dir/out.png" &
+    fi
+
+    # shellcheck disable=2086
+    gocr -u '?' $MODE \
+        -a ${CERTAINTY:-0} \
+        -c ${CHARSET:-'[0-9A-z]'} \
+        "$dir/out.png" 2>/dev/null | \
+    sed 's/ //g' | tr '[:lower:]' '[:upper:]'
+}
+
+while [ "$1" ] ; do
+    case $1 in
+        -h) usage ;;
+        -v) DEBUG=true ; shift ;;
+        -d) DISPLAY_RESULT=true ; shift ;;
+         *) break
+    esac
+done
+
+main "$@"

gotcha_v2

@@ -0,0 +1,110 @@
+#!/bin/sh
+#
+# Breaks simple captchas.
+#
+# deps: gocr, imagemagick, sed
+#
+
+# -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
+# CHARSET="[0-9A-z]"
+#### dataset2 only has numbers
+CHARSET="[0-9]"
+
+# 0-100, higher numbers will force strict matches
+CERTAINTY=0
+
+# 16: not dot divide overlapping chars
+# 32: do not context correct
+MODE='-m 16 -m 32'
+
+IMAGE_PROG=mpvimg
+# -*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*-*
+
+usage() {
+    >&2 printf "Usage: %s [-d] [-v] [file]\n" "${0##*/}"
+    exit 1
+}
+
+init() {
+    random() { r=$(($(date +%s) / $$ - $$)) ; echo ${r#-} ; }
+    dir=/tmp/$(random) # dir to store tmp files
+    tmp=$dir/$(random) # image created
+    unset -f random
+
+    clean
+    mkdir -p "$dir"
+}
+
+clean() {
+    rm -r "${dir:-?}" 2>/dev/null ||:
+}
+
+check() {
+    if [ ! -f "$1" ] ; then
+        >&2 echo "No such image: $1"
+        exit 1
+    fi
+}
+
+jpg2png() {
+    case $(file "$1") in
+        *jpg*|*JPG*|*jpeg*|*JPEG*)
+            convert "$1" "${1%.*}.png"
+            # rm "$1"
+            image=${image%.*}.png
+            ;;
+    esac
+}
+
+main() {
+    image=$1
+    check "$image"
+
+    init
+    # trap 'clean' INT TERM EXIT
+
+    # convert to png if necessary
+    jpg2png "$image"
+
+    # trim
+    convert -trim "$image" "$image"
+
+    # clean up text and display black on white background
+    # text will be evenly spaced but not aligned
+    convert \
+        -quality 100 \
+        -shave 1x1 \
+        -resize 500%x500% \
+        -threshold 50% \
+        -deskew 50 \
+        -trim \
+        -bordercolor white \
+        -border 30 \
+        "$image" "$tmp"
+
+    [ "$DEBUG" = true ] && $IMAGE_PROG "$tmp" &
+
+    mv -f "$tmp" "$dir"/out.png
+
+    if [ "$DEBUG" = true ] || [ "$DISPLAY_RESULT" = true ] ; then
+        $IMAGE_PROG "$dir/out.png" &
+    fi
+
+    # shellcheck disable=2086
+    gocr -u '?' $MODE \
+        -a ${CERTAINTY:-0} \
+        -c ${CHARSET:-'[0-9A-z]'} \
+        "$dir/out.png" 2>/dev/null | \
+    sed 's/ //g' | sed 's/\?/9/g' # very commonly '9' is '?'
+}
+
+while [ "$1" ] ; do
+    case $1 in
+        -h) usage ;;
+        -v) DEBUG=true ; shift ;;
+        -d) DISPLAY_RESULT=true ; shift ;;
+         *) break
+    esac
+done
+
+main "$@"

test.sh

@@ -0,0 +1,23 @@
+#!/bin/sh -e
+
+[ -d dataset ] || exit 1
+
+correct=0
+count=0
+total=$(printf '%s\n' dataset/* | wc -l)
+
+for i in dataset/*.png ; do
+    res=$(bash ./gotcha "$i")
+    i=${i#dataset/}
+    if [ "$res" = "${i%.png}" ] ; then
+        correct=$((correct + 1))
+        status=SUCCESS
+    else
+        status=FAIL
+    fi
+    printf '%s\n' "$status: [count: $count | correct: $correct/$total] - ANS: ${i%.png} RES: $res"
+    count=$((count + 1))
+done
+
+echo
+echo 'Done!'