StillHammer
a61a32b57f
- Move all Python scripts to tools/ directory - Move documentation files to docs/ directory - Create exams/ and homework/ directories for future use - Remove temporary test file (page1_preview.png) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude <noreply@anthropic.com>
106 lines
3.0 KiB
Python
106 lines
3.0 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Test different Y scaling factors to find the correct one.
|
|
The issue: center is aligned but top/bottom are off progressively.
|
|
This suggests the OCR coordinate space has a different Y scale than the image.
|
|
"""
|
|
|
|
import json
|
|
from pdf2image import convert_from_path
|
|
from reportlab.pdfgen import canvas
|
|
from reportlab.lib.utils import ImageReader
|
|
from reportlab.lib.colors import red, blue
|
|
from reportlab.pdfbase import pdfmetrics
|
|
from reportlab.pdfbase.ttfonts import TTFont
|
|
import tempfile
|
|
import os
|
|
|
|
# Font
|
|
try:
|
|
pdfmetrics.registerFont(TTFont('CF', 'C:/Windows/Fonts/msyh.ttc'))
|
|
font = 'CF'
|
|
except:
|
|
font = 'Helvetica'
|
|
|
|
# Load
|
|
with open('Raw/DevelopChinese/OCR/听力Intro.json') as f:
|
|
data = json.load(f)
|
|
|
|
# Convert
|
|
imgs = convert_from_path('Raw/DevelopChinese/PDF/听力Intro.pdf', dpi=300, first_page=1, last_page=1)
|
|
img = imgs[0]
|
|
w, h = img.size
|
|
print(f'Image: {w}x{h}')
|
|
|
|
# Find actual OCR max Y
|
|
details = data['pages'][0]['details']
|
|
max_y_ocr = max(max(p[1] for p in d['bbox']) for d in details if d.get('bbox'))
|
|
print(f'OCR max Y: {max_y_ocr}')
|
|
|
|
# Save image
|
|
with tempfile.NamedTemporaryFile(suffix='.png', delete=False) as tmp:
|
|
img.save(tmp.name, 'PNG')
|
|
tmp_path = tmp.name
|
|
|
|
# Create PDF
|
|
c = canvas.Canvas('test_Y_SCALING.pdf')
|
|
|
|
# Test different OCR heights (scale factors)
|
|
# If center is good but extremes are off, we need to find the right OCR height
|
|
test_ocr_heights = [3200, 3300, 3366, 3400, 3433, 3500, 3600, 3700]
|
|
|
|
for ocr_h in test_ocr_heights:
|
|
c.setPageSize((w, h))
|
|
c.drawImage(ImageReader(tmp_path), 0, 0, width=w, height=h)
|
|
|
|
# Draw title
|
|
c.setFont('Helvetica', 20)
|
|
c.setFillColorRGB(1, 0, 0)
|
|
scale = h / ocr_h
|
|
c.drawString(50, h - 50, f'OCR_HEIGHT: {ocr_h} (scale Y: {scale:.4f})')
|
|
|
|
# Draw boxes with this scaling
|
|
count = 0
|
|
for d in details:
|
|
text = d.get('text', '').strip()
|
|
if not text or count >= 30:
|
|
continue
|
|
bbox = d.get('bbox', [])
|
|
if not bbox or len(bbox) != 4:
|
|
continue
|
|
|
|
xs = [p[0] for p in bbox]
|
|
ys = [p[1] for p in bbox]
|
|
x = min(xs)
|
|
width = max(xs) - x
|
|
|
|
# Scale Y from OCR space (0-ocr_h) to PDF space (0-h)
|
|
y_ocr_max = max(ys)
|
|
y_ocr_min = min(ys)
|
|
|
|
# Map from OCR space to image space
|
|
y_pdf = h - ((y_ocr_max / ocr_h) * h)
|
|
height = ((y_ocr_max - y_ocr_min) / ocr_h) * h
|
|
|
|
c.setStrokeColor(red)
|
|
c.setLineWidth(0.5)
|
|
c.rect(x, y_pdf, width, height, stroke=1, fill=0)
|
|
|
|
font_size = max(1, height * 0.7)
|
|
c.setFont(font, font_size)
|
|
c.setFillColor(blue)
|
|
c.drawString(x, y_pdf, text)
|
|
|
|
count += 1
|
|
|
|
c.showPage()
|
|
print(f'Page with OCR height {ocr_h} done (scale: {scale:.4f})')
|
|
|
|
c.save()
|
|
os.unlink(tmp_path)
|
|
|
|
print(f'\nCreated test_Y_SCALING.pdf with {len(test_ocr_heights)} pages')
|
|
print('Find which OCR height gives best alignment everywhere!')
|
|
print(f'Current OCR max Y from data: {max_y_ocr}')
|
|
print(f'Image height: {h}')
|