Read barcodes from images in Python. The barKoder Python barcode scanner library decodes 1D and 2D barcodes from OpenCV and NumPy images, returns the results as JSON, and decodes locally on Linux, Windows and macOS with Python 3.11 or newer.
The barKoder Python SDK wraps the native barKoder decoding engine in a Python module. You load an image with OpenCV, pass the grayscale array to decodeImage, and get back JSON with every barcode found. Decoding runs on your machine, so it fits automation scripts, desktop tools and backend recognition pipelines. Read the September 2024 announcement of Python support.
python3 -m pip install opencv-python numpy
On Windows, the package also includes a wheel that you install with pip before you run the sample.
Start your free trialThis complete example is copied from the barkoder-python-sdk README. It reads qr.bmp, decodes QR and PDF417 codes, and prints each result.
from Barkoder import BarkoderSDK
import json
import numpy as np
import cv2 as cv
version = BarkoderSDK.getLibVersion();
print("SDK version: ", version)
status = BarkoderSDK.initialize("LICENSE_KEY");
print("SDK status: ", status)
#---------------configure the SDK--------------
QR = BarkoderSDK.constants["Decoders"]["QR"];
PDF417 = BarkoderSDK.constants["Decoders"]["PDF417"];
#set desired decoders
decoders = [QR, PDF417];
BarkoderSDK.setEnabledDecoders(decoders);
#set desired decodingSpeed
decodingSpeed = BarkoderSDK.constants["DecodingSpeed"]["Normal"];
BarkoderSDK.setDecodingSpeed(decodingSpeed);
#set desired ROI
BarkoderSDK.setRegionOfInterest(0, 0, 100, 100);
#set desired DPM mode
BarkoderSDK.setDatamatrixDpmModeEnabled(False);
#---------------configure the SDK--------------
#---------------read and scan an image---------
img = cv.imread('qr.bmp', cv.IMREAD_GRAYSCALE)
assert img is not None, "file could not be read, check with os.path.exists()"
height, width = img.shape[:2]
result_json = BarkoderSDK.decodeImage(img, width, height)
#---------------read and scan an image---------
#---------------handle the result--------------
result = json.loads(result_json)
print("resultsCount: ", result["resultsCount"])
# the result is a Python dictionary:
if result["resultsCount"] == 1:
print(result["barcodeTypeName"])
print(result["textualData"])
elif result["resultsCount"] > 1:
for res in result["results"]:
print(res["barcodeTypeName"])
print(res["textualData"])
print("---------------------------")
else:
print("No barcode found.")
#---------------handle the result--------------
Each result includes barcodeTypeName, textualData, binaryData, the four corner points of its location and a GS1 flag. See the Python API reference for every method and constant.
Python usually processes still images, not live camera streams. These settings from the Python docs help when you run the decoder over many files.
Set a maximum results count above 1 to get every barcode in an image. The docs example enables six barcode types, allows up to 10 results and uses slow decoding for thorough results.
#set multicode limit (value higher than 1 enables multicode)
BarkoderSDK.setMaximumResultsCount(10);
Rigorous may find more results but takes longer, and the API reference recommends it only when scanning from an image. That makes it a fit for batch jobs where time per file matters less than finding every code. The example below sets Slow, and the Rigorous constant works the same way.
decodingSpeed = BarkoderSDK.constants["DecodingSpeed"]["Slow"];
BarkoderSDK.setDecodingSpeed(decodingSpeed);
For hard samples, the docs show OpenCV steps that improve detection: unsharp masking and erosion for low-contrast QR codes, dilation for dotted Data Matrix codes, and finding, deskewing and cropping the barcode region.
# Apply unsharp mask with 500% sharpening
unsharp_image = cv.addWeighted(img, 5.0, gaussian_blur, -4.0, 0)
The Python SDK uses the same decoding engine as the other barKoder SDKs. Each feature is one setting.
Reads deformed and misshaped 1D barcodes. It is off by default, so enable it in your configuration.
enableMisshaped1D = BarkoderSDK.constants["EnableMisshaped1D"]["Enable"];
BarkoderSDK.setEnableMisshaped1D(enableMisshaped1D);
Enable only the DotCode decoder and set the decoding speed to slow.
Dotcode = BarkoderSDK.constants["Decoders"]["Dotcode"];
#set Dotcode decoder only
BarkoderSDK.setEnabledDecoders([Dotcode], 1);
decodingSpeed = BarkoderSDK.constants["DecodingSpeed"]["Slow"];
BarkoderSDK.setDecodingSpeed(decodingSpeed);
Reads EAN and UPC codes with motion, lens or focus blur.
upcEanDeblur = BarkoderSDK.constants["UpcEanDeblur"]["Enable"];
BarkoderSDK.setUpcEanDeblur(upcEanDeblur);
2D codes
QR Code, Micro QR, PDF417, Micro PDF417, Data Matrix, Aztec, Aztec Compact, MaxiCode and DotCode.
1D, industrial and postal codes
Code 128, Code 93, Code 39, Code 32, Codabar, Code 11, MSI, UPC-A, UPC-E, EAN-13, EAN-8, Code 25, Interleaved 2 of 5, ITF-14, IATA 2 of 5, Matrix 2 of 5, Datalogic 2 of 5, COOP 2 of 5, Telepen, GS1 DataBar, Australian Post, Royal Mail, KIX, Japanese Post, PostNet, Planet and IMB.
Open-source Python barcode libraries are a good start. Here is where they stop and when a commercial SDK is worth testing.
Run each library over the same folder of your real images and compare how many barcodes each one decodes.
Typical Python projects for the SDK, from the barKoder Python product page.
Verified credentials
Certified information security management built around internationally recognized controls.
Official Solution Provider supporting standards-based barcode and data-capture workflows.
Solution Provider listing covering barcode, mobile application and 2D barcode capabilities.
Discover the barKoder Web SDK listing and connect through Microsoft Marketplace.
Registered Independent Software Vendor in Zebra's PartnerConnect ecosystem.
The engine behind the Python SDK also ships as native, web, desktop and cross-platform SDKs, so the barcodes your pipeline reads can also be scanned in your mobile and web apps.
Direct SDKs for mobile, browser, desktop and server environments.
Official integrations for teams building one application across platforms.
Install the free barKoder demo app, or open the web demo, and scan your own barcodes before you integrate the SDK.
iPhone and iPad
Scan with your camera, or point it at the QR code to open the App Store.
Android phones and tablets
Scan with your camera, or point it at the QR code to open Google Play.
Any modern browser
Runs the Web SDK in your browser, on desktop or mobile, with nothing to install.
Test with known samples. Print or display the testing sheet and scan each 1D and 2D barcode with the demo app. The data sheet summarizes the SDK for your team.
Initialize the SDK with your license key, enable the barcode types you need, then load the image as grayscale with cv.imread(path, cv.IMREAD_GRAYSCALE) and pass it to BarkoderSDK.decodeImage(img, width, height). The call returns JSON. Parse it with json.loads and read resultsCount, barcodeTypeName and textualData. The Python general example shows each step.
For clean QR codes and common 1D barcodes, pyzbar or zxing-cpp can be enough. If your images include Data Matrix, PDF417, DotCode or postal codes, damaged or blurred barcodes, or several codes per image, test barKoder on the same folder of images and compare how many barcodes each library decodes.
Python 3.11 or newer. The barKoder docs cover 64-bit Linux on x86-64, with Ubuntu 22.04 or newer recommended. The barkoder-python-sdk README also provides Linux ARM64, Windows 10 and 11 on x86-64, and macOS 11 or newer for Intel and Apple silicon. Contact barKoder if you need an older Python version.
No. Download the package from the barKoder Developer Portal or clone the GitHub repository. You install opencv-python and numpy with pip, and on Windows you install the SDK wheel included in the package with pip.
Yes. Call BarkoderSDK.setMaximumResultsCount with a value above 1. When more than one barcode is found, the JSON contains a results list with the type, text, raw bytes and location of each barcode.
decodeImage takes a grayscale NumPy array, so frames you capture with OpenCV go through the same call. The API reference notes that the multicode caching duration setting is typically useful when scanning from a live camera stream. Most Python projects decode still images.
Decoding runs offline on your machine. The SDK connects only for the license check. Each machine must reach the licensing server at least once every 30 days to keep returning licensed results.
Generate a key on the trial page or in the barKoder Developer Portal. A trial license lasts 30 days and covers up to 25 devices, with unlimited scans. No credit card is required. For server deployments, see the custom terms on the pricing page.
Run the barKoder Python SDK on your own images with a free 30-day trial, or talk to an SDK expert about your pipeline.
