deeplearning.ai-Foundations-of-Convolutional-Neural-Networks

Foundations of Convolutional Neural Networks, deeplearning.ai coursera course

Week 1: Foundations of Convolutional Neural Networks

Introduction to Convolution, pooling and paddnig.

• Familiar formula: conv layer output size = (n + 2*p - k)/s + 1

• Number of params in ten 3x3x3 filtres: (3x3x3 + 1[bais]) x 10 = 280

Assignment 1: Implement conv layer in numpy (forward/backward)

Assignment 2: Intro. to TensoeFlow

Week 2: Deep convolutional models: case studies

ResNet

• Residual block: [image to be updated]
• Why it works? Skip-connection make NNs easy to learn. Prevent gradient vanishing.

Inception

• Bottleneck layer: Apply 1x1 conv to shrink channle size
• Concatenate output of diffeernt conv routes

Assignment 1: Intro. to Keras

• Input(shape=...) => [Conv2D/BN/ReLU] x N => model=Model(input, output) => mdoel.compile(...) => model.fit(...)

Assignmnet 2: ResNet50

• convolution_block: A block that reduces dimnesion by 2 using stride2 Conv2D
• identity_block: Con2D/BN/ReLU => Conv2D/BN => Add => ReLU  

Week 3: Object detection

Week 4: Special applications: Face recognition & Neural style transfer