Python OCR Tutorial: Recognize SEMI Semiconductor Fonts with Dynamsoft Capture Vision SDK

SEMI (Semiconductor Equipment and Materials International) font is a special dot matrix font used for marking silicon wafers. In this tutorial, we’ll walk through building a Python application to recognize these specialized markings using Dynamsoft Capture Vision SDK.

What you’ll build: A cross-platform Python script that loads a custom SEMI OCR model, processes semiconductor wafer images, and overlays recognized dot-matrix font characters as bounding-box annotations using Dynamsoft Capture Vision SDK.

Key Takeaways

  • Dynamsoft Capture Vision SDK supports custom OCR models, enabling recognition of non-standard fonts such as the SEMI dot-matrix font used on silicon wafers.
  • The CaptureVisionRouter class accepts user-supplied model buffers via append_model_buffer(), making the pipeline extensible to any industry-specific character set.
  • Recognition results include per-character bounding-box location data, making it straightforward to overlay text annotations on wafer images using OpenCV.
  • The complete solution runs on Windows, Linux, and macOS with a single pip install command.

Common Developer Questions

  • How do I recognize SEMI font characters on semiconductor wafers using Python OCR?
  • Can Dynamsoft Capture Vision SDK read custom or non-standard dot-matrix fonts?
  • How do I load a custom OCR model into Dynamsoft Capture Vision SDK in Python?

Watch SEMI OCR Font Recognition in Action

Prerequisites

  • Python 3.8 or later
  • Dynamsoft Capture Vision Trial License: Get a 30-Day trial license key for the Dynamsoft Capture Vision SDK.

  • Python Packages: Install the required Python packages using the following commands:

      pip install dynamsoft-capture-vision-bundle opencv-python
    • dynamsoft-capture-vision-bundle: Python binding for Dynamsoft Capture Vision SDK.
    • opencv-python: For displaying source images and overlaying recognition results.

What This Implementation Provides

  • Specialized SEMI Font Recognition: Uses a custom model trained for single-density dot matrix fonts (uppercase letters A-Z and digits 0-9).
  • Visual Feedback: Draws bounding boxes around recognized text.
  • Batch Processing: Processes single images or entire directories.
  • Cross-Platform: Works on Windows, Linux, and macOS.

Semi-Font OCR Recognition

Step 1: Initialize the Dynamsoft Capture Vision SDK

Create a new Python file and initialize the SDK with your license key:

from dynamsoft_capture_vision_bundle import *

err_code, err_str = LicenseManager.init_license("LICENSE-KEY")
if err_code != EnumErrorCode.EC_OK and err_code != EnumErrorCode.EC_LICENSE_CACHE_USED:
    print("License initialization failed: " + err_str)

Step 2: Load the SEMI OCR Model

A custom model trained by Dynamsoft enables the Capture Vision SDK to recognize SEMI fonts:

cvr = CaptureVisionRouter()

# Load the SEMI OCR model
with open('models/semi-ocr.data', 'rb') as f:
    model_data = f.read()
    
err_code, err_str = cvr.append_model_buffer('semi-ocr', model_data, 1)
if err_code != EnumErrorCode.EC_OK:
    print("Model loading failed: " + err_str)

For model-related questions, please contact Dynamsoft Support.

Step 3: Configure SEMI OCR Recognition from a JSON Settings File

Besides the model file, recognition settings must be loaded from a semi-ocr.json file.

err_code, err_str = cvr.init_settings_from_file("semi-ocr.json")
if err_code != EnumErrorCode.EC_OK:
    print("Configuration loading failed: " + err_str)

Step 4: Process Wafer Images and Visualize Recognition Results

Here’s the core recognition logic that processes images and overlays results:

import cv2
import numpy as np

def process_image(image_path, cvr):
    cv_image = cv2.imread(image_path)
                    
    result = cvr.capture(image_path, "recognize_semi_ocr")
    
    if result.get_error_code() != EnumErrorCode.EC_OK:
        print("Error: " + str(result.get_error_code())+ result.get_error_string())
    else:
        items = result.get_items()
        for item in items:
            if isinstance(item, TextLineResultItem):
                print(f"{RED}{item.get_text()}{RESET}")

                location = item.get_location()
                points = [(p.x, p.y) for p in location.points]
                cv2.drawContours(cv_image, [np.intp(points)], 0, (0, 255, 0), 2)

                cv2.putText(cv_image, item.get_text(), (points[0][0] + 10, points[0][1] + 20),
        
        cv2.FONT_HERSHEY_SIMPLEX, 0.5, (0, 0, 255), 1)
    
        cv2.imshow(
            os.path.basename(image_path), cv_image)

Step 5: Build an Interactive File and Directory Scanner

Add a loop to handle single files or directories:

import os

def main():
    cvr = CaptureVisionRouter()
    # ... initialization code from previous steps ...
    
    while True:
        path = input("Enter image path or directory (Q to quit): ").strip()
        
        if path.lower() == "q":
            break
            
        if not os.path.exists(path):
            print("File not found: " + path)
            continue
        else:
            if os.path.isfile(path):
                process_image(path, cvr)
            elif os.path.isdir(path):
                files = os.listdir(path)
                for file in files:
                    if file.endswith(".jpg") or file.endswith(".jpeg") or file.endswith(".png"):
                        process_image(os.path.join(path, file), cvr)
                    
            cv2.waitKey(0)
            cv2.destroyAllWindows()

if __name__ == '__main__':
    main()

Step 6: Run the SEMI Font OCR Application

python read_semi_ocr.py

SEMI OCR Font from multiple images

Common Issues & Edge Cases

  • Model file not found: Ensure models/semi-ocr.data is present in the working directory. append_model_buffer() returns a non-zero error code if the file cannot be read — inspect err_str for details.
  • EC_LICENSE_CACHE_USED on startup: This is not a failure; the SDK is using a locally cached license and recognition proceeds normally. Only treat the result as an error when the code is neither EC_OK nor EC_LICENSE_CACHE_USED.
  • Empty results from get_items(): Verify that semi-ocr.json was loaded successfully and that the task name passed to cvr.capture() — "recognize_semi_ocr" — exactly matches the template name defined in that JSON file.
  • Low accuracy on degraded wafers: The custom model targets clean single-density dot-matrix fonts. For heavily worn or etched surfaces, apply contrast enhancement or grayscale normalization with OpenCV before calling cvr.capture().

Source Code

https://github.com/yushulx/python-barcode-qrcode-sdk/edit/main/examples/official/semi_font_ocr