Last modified on 27 Jul 2020.
This is my note for the course (Convolutional Neural Networks). The codes in this note are rewritten to be more clear and concise.
π― Overview of all 5 courses.
π Course 1 β Neural Networks and Deep Learning.
π Course 2 β Improving Deep Neural Networks: Hyperparameter tuning, Regularization and Optimization.
π Course 3 β Structuring Machine Learning Projects.
π Course 5 β Sequence Models.
This course will teach you how to build convolutional neural networks and apply it to image data.
Computer vision
 Challenges of computer vision: inputs get really big. $\Rightarrow$ Easily lead to overfitting and infeasible!
Edge Detection

We wanna detect vertical / horizontal edges. $*$ is called a convolution operator. The $3\times 3$ square is called a filter (kernel)
Horizontal edge detection.^{ref}
Vertical vs Horizontal.^{ref}

There are many different filers we could use. Sobel filter and Scharr filter βs advantage is that it allows you to put a little bit more weight to the central row of the central pixel, this makes it maybe a little bit more robust.
Sobel filter and Scharr filter.^{ref}
References
π Course 5 β Sequence Models.
^{β’}Notes with this notation aren't good enough. They are being updated. If you can see this, you are so smart. ;)