CNNs learn features from raw pixels: input feature map (width, height, RGB depth) and a stack of convolution, ReLU, pooling.
Intro to CNN.
https://developers.google.com/machine-learning/practica/image-classification/convolutional-neural-networks
Convoluted Neural Network. Breakthrough: progressively extract higher- and higher-level representations of the image content.
Instead of preprocessing the data to derive features like textures and shapes, a CNN takes just the image's raw pixel data as input and "learns" how to extract these features, and ultimately infer what object they constitute.
Input feature map
Matrix:
- Width: image width.
- Height: image height.
- Depth: R, G, B.
Modules
A CNN comprises a stack of modules. Each module performs 3 tasks:
- Convolution.
- ReLU.
- Pooling.
Convoluted map
Convolution extracts tiles of the input feature map and applies filters. Two factors:
- Size of tile — typically 3x3 or 5x5.
- Depth: number of filters.
Application of filter:
