Skip to main content

AlexNet

AlexNet is a deep convolutional neural network (CNN) architecture that was introduced by Alex Krizhevsky, Ilya Sutskever, and Geoffrey Hinton in 2012. It achieved significant success in the ImageNet Large Scale Visual Recognition Challenge (ILSVRC) and played a pivotal role in popularizing deep learning for computer vision tasks.

Architecture:

  • Input: 224x224 RGB image
  • C1: Convolutional layer with filters of size 11x11, followed by a ReLU activation and max pooling.
  • N1: Local response normalization layer to normalize the activations.
  • C2: Convolutional layer with filters of size 5x5, followed by a ReLU activation and max pooling.
  • N2: Local response normalization layer to normalize the activations.
  • C3: Convolutional layer with filters of size 3x3, followed by a ReLU activation.
  • C4: Convolutional layer with 384 filters of size 3x3, followed by a ReLU activation.
  • C5: Convolutional layer with 256 filters of size 3x3, followed by a ReLU activation and max pooling.
  • F6: Fully connected layer with 4096 neurons, followed by a ReLU activation and dropout.
  • F7: Fully connected layer with 4096 neurons, followed by a ReLU activation and dropout.
  • Output: Fully connected layer with 1000 neurons corresponding to the 1000 classes in the ImageNet dataset, followed by a softmax activation for classification.