AlexNet-Visualization

Visualizing layer output of AlexNet model trained on cifar-10 dataset

Architecture

Official Paper

Model Summary

Trained on cifar-10 dataset

Performance

loss: 0.5992 - accuracy: 0.8075 - val_loss: 1.3247 - val_accuracy: 0.7012(Trained for 18 epochs)

Summary

Visualization

Original Image

Layer-1(conv2D)

Below are first five filters of Layer-1 resized to (224,224) pixels

Layer-4(conv2D)

Below are first five filters of Layer-4 resized to (224,224) pixels

Layer-7(conv2D)

Below are random five filters of Layer-7 resized to (224,224) pixels

Layer-11(conv2D)

Below are first five filters of Layer-11 resized to (224,224) pixels

Try it yourself on different images on google colab or on local system by using checkpoint and AlexNet.ipynb in the repository