Chapter One
Computer Vision core issues, how to ignore differences inside the same body, and strengthen the distinction between different objects.
Artificial neural networks.
Back-propagation algorithm. Complex dismantling of the chain rule is independent of the context of the connection layer, in accordance with the respective weights assigned incorrectly updated. Such an event is not positioned to make predictions of the existing data by statistical laws.
In 2006, training deep neural networks breakthrough. Using more and more hidden layer neurons have a better ability to learn.
CNN: modeled on biological visual layer decomposition algorithm.
Training platform, using the model, speed and cycle.
Commonly used TensorFlow / Cafe / PyTroch
The core task processing object, comprising a detection, identification, segmentation, feature points, sequence learning.
Chapter two
Anoconda