Data Science — Computer Vision

Image loading

import cv2
import matplotlib.pyplot as plt

# Load image
img = cv2.imread('image.jpg')
img_rgb = cv2.cvtColor(img, cv2.COLOR_BGR2RGB)

plt.imshow(img_rgb)
plt.show()

Image processing

# Resize
img_resized = cv2.resize(img, (224, 224))

# Grayscale
gray = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)

# Blur
blurred = cv2.GaussianBlur(img, (5, 5), 0)

# Edge detection
edges = cv2.Canny(gray, 100, 200)

Object detection

import tensorflow as tf

# Load pre-trained model
model = tf.keras.applications.MobileNetV2(weights='imagenet')

# Predict
predictions = model.predict(img_preprocessed)

Transfer learning

base_model = tf.keras.applications.VGG16(weights='imagenet', include_top=False)
base_model.trainable = False

model = keras.Sequential([
    base_model,
    keras.layers.GlobalAveragePooling2D(),
    keras.layers.Dense(10, activation='softmax')
])

Mini Practice

1. Load and display images
2. Apply filters
3. Detect edges
4. Use transfer learning

Up Next

Continue with Recommender Systems - Building recommendation engines.