a4ocr

#!/bin/bash

function usage() {
  cat <<EOF
OCR of a scan of a A4 image as markdown using chatgpt
Usage: $(basename $0) IMAGE
EOF
  exit 1
}

inputimage="$1"

if [ "$#" -eq 0 ] || [ "$1" = "--help" ] || [ "$1" = "-h" ] || [ -z "$inputimage" ]; then
  usage
fi

width=$(identify -format "%w" "$inputimage")
height=$(identify -format "%h" "$inputimage")

echo "Image dimensions: $width x $height"
echo "Should be A4 portrait mode!"

# check that height / width is between 1.4 and 1.5
ratio=$(echo "scale=2; $height / $width" | bc)
if [ $(echo "$ratio > 1.5" | bc) -eq 1 ] || [ $(echo "$ratio < 1.4" | bc) -eq 1 ]; then
    echo "The image does not have the correct aspect ratio for A4 portrait mode - it has a ratio of $ratio"
    exit 1
fi

# cut the image into 3 stripes, format 1999x767

stripe_height=$((width * 767 / 1999))
shift=$(( ($height - $stripe_height) / 4 + 1))

for i in {0..4};
do
    top=$((i * shift))
    if [ "$top" -lt 0 ]; then top=0; fi

    bottom=$((i * shift + stripe_height))
    if [ "$bottom" -gt "$height" ]; then bottom=$height; fi

    output_image="${inputimage}_stripe$((i + 1)).png"
    magick "$inputimage" -crop "${width}x$((bottom - top))+$((0))+$top" +repage "$output_image"
    echo "Saved crop ${width}x$((bottom - top))+$((0))+$top to $output_image"
done

for fil in ${inputimage}_stripe*.png; do
  chatgpt -i "$fil" -id high "Output the visible part of a text as raw markdown, without any additional comment and without marking it as code block. IMPORTANT: Do not add or remove anything from the text!"
  echo "============================================================"
done > "${inputimage}_transcribed_raw.md"

# output file without extension, which might be .jpg, .png and so forth
outputfil="${inputimage%%.*}"

(cat <<EOF
The input contains transcriptions of 5 parts of a text, which overlap somewhat. The parts
of the text are separated with "============================================================" .
Join the parts of the text together so that the overlaps are removed
and output the result as raw markdown, without any comments and without marking it as code block.
EOF
) | chatgpt -f "${inputimage}_transcribed_raw.md" - > ${outputfil}.md

/bin/rm -f "${inputimage}_stripe*.png"
/bin/rm -f "${inputimage}_transcribed_raw.md"
echo "Removed temporary files including ${inputimage}_transcribed_raw.md"

echo "Wrote recognized text to ${outputfil}.md"