This sources gives a basic overview of convolutional neural networks and their application as the computer vision component of self driving vehicles. The article builds a mental model of the netork and breaks down the design into steps; Input image, Hidden layers (convolutional layers, pooling layers, and activation layers), and an Output layer which is a probability distribution.
Key Points
Uber and Tesla use convolutional neural networks as the computer vision components of their self driving vehicles.
CNNs are a multi-layered feed-forward neural network.
Used to process structured arrays of data(ex. images).
More convolutional layers = more complex pattern recognition.
Convolutional layers are made of a series of convolutional kernels which convolve across the image looking for patterns. The patterns are not programed and are instead learned by the NN during training.
After passing the image through a convolutional layer the output is passed through an activation function. Common ones include sigmoid function, ReLU function.
Title
Convolutional Neural Networks
URL
https://deepai.org/machine-learning-glossary-and-terms/convolutional-neural-network
Summary
This sources gives a basic overview of convolutional neural networks and their application as the computer vision component of self driving vehicles. The article builds a mental model of the netork and breaks down the design into steps; Input image, Hidden layers (convolutional layers, pooling layers, and activation layers), and an Output layer which is a probability distribution.
Key Points
Uber and Tesla use convolutional neural networks as the computer vision components of their self driving vehicles. CNNs are a multi-layered feed-forward neural network. Used to process structured arrays of data(ex. images). More convolutional layers = more complex pattern recognition. Convolutional layers are made of a series of convolutional kernels which convolve across the image looking for patterns. The patterns are not programed and are instead learned by the NN during training. After passing the image through a convolutional layer the output is passed through an activation function. Common ones include sigmoid function, ReLU function.
Citation
T. Wood, “Convolutional Neural Networks,” DeepAI, May 17, 2019. https://deepai.org/machine-learning-glossary-and-terms/convolutional-neural-network
Repo link