draw_ocr inhomogeneous shape after 1 dimensions

[emadyounan]

🔎 Search before asking

• I have searched the PaddleOCR Docs and found no similar bug report.
• I have searched the PaddleOCR Issues and found no similar bug report.
• I have searched the PaddleOCR Discussions and found no similar bug report.

🐛 Bug (问题描述)

Hi everyone, I'm new with PaddleOcr and I was testing it with my image

I used this code to OCR the Image

Perform OCR on the image

image_path = "/content/XRN04250159554A-001Page_1.tif"
img = cv2.imread(image_path)
gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

results = ocr.ocr(gray_image, cls=True)

# Extract and print the detected text
for result in results:
    if result:
        for line in result:
            print(f"Detected Text: {line[1][0]} | Confidence: {line[1][1]}")

I got this result

[2025/03/05 17:22:47] ppocr DEBUG: dt_boxes num : 1, elapsed : 0.02038741111755371
[2025/03/05 17:22:47] ppocr DEBUG: rec_res num  : 1, elapsed : 0.11077070236206055
Detected Text: 30193469 | Confidence: 0.9998676180839539

But I need to show the image with showing OCR data into image
I tried this one

# draw result
from PIL import Image, ImageDraw, ImageFont
image = Image.open(image_path).convert('RGB')

boxes = [line[0] for line in result]
txts = [line[0][1][0] for line in result]
scores = [line[0][1][1] for line in result]

font = ImageFont.load_default()
im_show = draw_ocr(image, boxes, txts, scores,font_path=font)
im_show = Image.fromarray(im_show)
im_show.save('result.jpg')

But I got this error

ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions. The detected shape was (2,) + inhomogeneous part.

🏃‍♂️ Environment (运行环境)

I'm using Google Colab

🌰 Minimal Reproducible Example (最小可复现问题的Demo)

That's example image

[emadyounan]

I'm using last version of PaddleOCR

[GreatV]

The error "ValueError: setting an array element with a sequence. The requested array has an inhomogeneous shape after 1 dimensions." suggests that there is an issue with how the data is structured before passing it to draw_ocr.

🔍 Possible Issues and Solutions:

1. Incorrect List Indexing

Your code contains:

boxes = [line[0] for line in result]
txts = [line[0][1][0] for line in result]  # Incorrect indexing
scores = [line[0][1][1] for line in result]  # Incorrect indexing

Here, line[0][1][0] is incorrect because the line structure inside results looks like this:

[
    [ [[x1, y1], [x2, y2], [x3, y3], [x4, y4]], ('Detected Text', Confidence_Score) ]
]

Fix it by correcting the indexing:

txts = [line[1][0] for line in result]  # Extracting the detected text
scores = [line[1][1] for line in result]  # Extracting confidence score

2. Missing Font Path

In your code:

font = ImageFont.load_default()
im_show = draw_ocr(image, boxes, txts, scores, font_path=font)

The font_path argument requires a string path to a valid font file (like .ttf), but ImageFont.load_default() does not return a valid file path.

Fix it by providing a proper font path:

im_show = draw_ocr(image, boxes, txts, scores, font_path='/usr/share/fonts/truetype/dejavu/DejaVuSans.ttf')

If you're using Google Colab, install a font like DejaVu first:

!apt-get install fonts-dejavu
font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"

3. Ensure result is Processed Correctly

Your code declares:

results = ocr.ocr(gray_image, cls=True)

PaddleOCR returns a list of lists, so you need to ensure consistent processing:

if results and results[0]:  # Ensure results are non-empty
    result = results[0]  # Extract the first list of detections

Then, apply the correct data extraction:

boxes = [line[0] for line in result]
txts = [line[1][0] for line in result]
scores = [line[1][1] for line in result]

---

✅ Final Fixed Code

from paddleocr import PaddleOCR, draw_ocr
from PIL import Image, ImageDraw, ImageFont
import cv2

# Load PaddleOCR
ocr = PaddleOCR(lang='en')

# Read and preprocess the image
image_path = "/content/XRN04250159554A-001Page_1.tif"
img = cv2.imread(image_path)
gray_image = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# Perform OCR
results = ocr.ocr(gray_image, cls=True)

# Check we have detections
if results and results[0]:  
    result = results[0]
    
    # Extract bounding boxes, text, and confidence scores
    boxes = [line[0] for line in result]
    txts = [line[1][0] for line in result]
    scores = [line[1][1] for line in result]

    # Open image
    image = Image.open(image_path).convert('RGB')

    # Specify font path
    font_path = "/usr/share/fonts/truetype/dejavu/DejaVuSans-Bold.ttf"  # Adjust based on the system

    # Draw OCR results on image
    im_show = draw_ocr(image, boxes, txts, scores, font_path=font_path)
    im_show = Image.fromarray(im_show)
    im_show.save('result.jpg')
    print("Result image saved successfully.")
else:
    print("No text detected.")

🎯 Summary of Fixes

1. Fixed incorrect list indexing when retrieving text and confidence scores.
2. Provided a valid font path instead of using ImageFont.load_default().
3. Handled empty OCR results safely by checking if results and results[0].
4. Formatted the code properly to improve readability.

Now, your PaddleOCR script should correctly annotate an image with detected text without raising errors. 🚀

---

Response generated by 🤖 feifei-bot | chatgpt-4o-latest

⚠️ This is an AI-generated response. Please verify any critical information.