DL : CNN & Computer vision

Part 4.2 of Deep Learning Specialization

1. Classification vs Localization vs Detection

1.1 Object Localization

boundary center (bx, by)

boundary frame (bw, bh)

1.2 Loss Function

- p = 1 (object in image) : Sum(predict-actual)²

- p = 0 (no object in image) : (prob.predict — prob.actual)²

1.3 Anchor box for Overlapping objects

For each sliding window,

you predict N Anchor boxes for N class of objects

1.4 Object Detection — YOLO algorithm (You Look Only Once)

1. Slide image into small windows

2. Predict class & boundary box of each window

Classes of objects found in each window

Boundary of objects found in each window

3. Non-Max Suppression

- Remove low possible classes of objects

- Remove boundary boxes of the same objects

2. Face Verification vs Face Recognition

3. Neural Style Transfer

Content image + Style image = Generated image

Loss Function = (Wc* Lcontent) + (Ws * Lstyle)

Related Articles

• Neural Network — Part 1.1
• Neural Network — Part 1.2
• Neural Network — Part 2
• Neural Network — Part 3
• CNN — Part 4.1
• CNN — Part 4.2
• RNN — Part 5.1
• RNN — Part 5.2

Reference

Deep Learning Specialization: Convolutional Neural Network (Coursera) (Youtube)