← Deep Learning learning pathDeep Learning · Module 6 of 10
Convolutional neural networks
Process images with CNNs.
📖 Key ideas
- Convolutions slide small filters over images to detect features.
- Early layers detect edges; deeper layers detect shapes and objects.
- Pooling reduces spatial size.
- CNNs share weights, needing fewer parameters.
- Famous architectures: LeNet, AlexNet, VGG, ResNet.
💡 Real-life example: A convolution filter is a magnifying glass sliding across a photo looking for a specific pattern.
ConvolutionFilter sliding over input
Filter/kernelSmall learned weight matrix
PoolingDownsampling feature maps
Feature mapOutput of a convolution