OpenCV Python Digit Recognition with KNN Tutorial: Handwritten Digit Classification Guide

OpenCV Python – Digit Recognition with KNN

Digit recognition is one of the most popular machine learning applications in computer vision. It involves identifying handwritten or printed digits from images.

OpenCV provides a built-in implementation of the K-Nearest Neighbors (KNN) algorithm, making it easy to build a digit recognition system.

In this tutorial, you'll learn how to recognize handwritten digits using OpenCV Python and KNN.

---

1. What is Digit Recognition?

Digit recognition is the process of automatically identifying numerical digits such as:

0 1 2 3 4 5 6 7 8 9

Applications include:

• Postal code recognition
• Bank cheque processing
• OCR systems
• Document digitization
• Number plate recognition
• Educational software

---

2. What is KNN?

K-Nearest Neighbors (KNN) is a supervised machine learning algorithm.

It classifies data by:

1. Finding the nearest training samples
2. Looking at their labels
3. Predicting the most common label

Example:

Neighbors:
3, 3, 3, 8, 3

Prediction:
3

---

3. Import Required Libraries

import cv2
import numpy as np

---

4. Understanding the Dataset

OpenCV provides a sample digit dataset containing handwritten digits.

The dataset image contains:

5000 digits
50 rows
100 columns
Digits 0–9

Each digit is represented by:

20 × 20 pixels

---

5. Load the Digits Dataset

img = cv2.imread('digits.png', 0)

cv2.imshow("Digits Dataset", img)
cv2.waitKey(0)
cv2.destroyAllWindows()

---

6. Split Dataset into Individual Digits

cells = [np.hsplit(row,100) for row in np.vsplit(img,50)]

x = np.array(cells)

Output shape:

print(x.shape)

Output:

(50, 100, 20, 20)

---

7. Prepare Training and Testing Data

Convert images into feature vectors.

train = x[:,:50].reshape(-1,400).astype(np.float32)

test = x[:,50:100].reshape(-1,400).astype(np.float32)

Explanation:

20 × 20 = 400 pixels

Each digit becomes:
[400 features]

---

8. Create Labels

Generate labels for digits 0–9.

k = np.arange(10)

train_labels = np.repeat(k,250)[:,np.newaxis]
test_labels = np.repeat(k,250)[:,np.newaxis]

---

9. Create KNN Model

OpenCV provides the KNN classifier.

knn = cv2.ml.KNearest_create()

---

10. Train the KNN Model

knn.train(
    train,
    cv2.ml.ROW_SAMPLE,
    train_labels
)

Training is now complete.

---

11. Test the Model

Use K = 5 neighbors.

ret, result, neighbours, dist = knn.findNearest(
    test,
    k=5
)

---

12. Calculate Accuracy

matches = result == test_labels

correct = np.count_nonzero(matches)

accuracy = correct * 100.0 / result.size

print("Accuracy:", accuracy)

Example Output:

Accuracy: 91.2

Accuracy may vary depending on the dataset.

---

13. Recognize a New Digit

Load a new handwritten digit image.

digit = cv2.imread("digit.png", 0)

digit = cv2.resize(digit, (20,20))

sample = digit.reshape(-1,400).astype(np.float32)

Predict:

ret, result, neighbours, dist = knn.findNearest(
    sample,
    k=5
)

print("Predicted Digit:", int(result[0][0]))

Output:

Predicted Digit: 7

---

14. Understanding K Value

The K value determines how many neighbors are considered.

Example:

K Value	Behavior
1	Fast but sensitive to noise
3	Common choice
5	Balanced
7+	More stable but slower

Typical choice:

k = 3 or 5

---

15. Advantages of KNN

Simple to Implement

No complex training process.

Good Accuracy

Works well on small datasets.

Easy to Understand

Ideal for beginners learning machine learning.

Flexible

Can classify various types of data.

---

16. Limitations of KNN

Slow on Large Datasets

Every prediction requires distance calculations.

Memory Intensive

Stores all training samples.

Sensitive to Noise

Incorrect samples may affect predictions.

---

17. Real-World Applications

Digit recognition with KNN is used in:

• OCR systems
• Bank cheque readers
• Handwritten form processing
• Educational apps
• Automated data entry
• Postal sorting systems

---

18. Complete Example

import cv2
import numpy as np

img = cv2.imread('digits.png',0)

cells = [np.hsplit(row,100) for row in np.vsplit(img,50)]
x = np.array(cells)

train = x[:,:50].reshape(-1,400).astype(np.float32)
test = x[:,50:100].reshape(-1,400).astype(np.float32)

k = np.arange(10)

train_labels = np.repeat(k,250)[:,np.newaxis]
test_labels = np.repeat(k,250)[:,np.newaxis]

knn = cv2.ml.KNearest_create()

knn.train(
    train,
    cv2.ml.ROW_SAMPLE,
    train_labels
)

ret, result, neighbours, dist = knn.findNearest(
    test,
    k=5
)

matches = result == test_labels

correct = np.count_nonzero(matches)

accuracy = correct * 100.0 / result.size

print("Accuracy:", accuracy)

---

19. Best Practices

• Normalize image sizes before training
• Use grayscale images
• Choose appropriate K values
• Remove noisy samples
• Evaluate accuracy using test data
• Use larger datasets for better performance

---

Conclusion

Digit Recognition with KNN in OpenCV Python is an excellent introduction to machine learning and computer vision. By combining image preprocessing with the K-Nearest Neighbors algorithm, you can build systems capable of recognizing handwritten digits with impressive accuracy.

After mastering KNN digit recognition, you can explore more advanced techniques such as Support Vector Machines (SVM), Artificial Neural Networks (ANN), and Deep Learning models like CNNs for even higher accuracy.