UpSampling2D: how to use upsampling with Keras?
The Convolutional layers section of the Keras API contains the so-called UpSampling2D layer. But what does it do? And how...
Conv2DTranspose: using 2D transposed convolutions with Keras
Transposed convolutions – we looked at them in theory in a previous blog post, but how can they be applied?...
How to use ELU with Keras?
Exponential Linear Unit is a new type of activation function that attempts to resolve both the defects of traditional ReLU...
How to use PReLU with Keras?
Rectified Linear Unit, or ReLU, is considered to be the standard activation function of choice for today’s neural networks. Even...
Visualize Keras models: overview of visualization methods & tools
Visualizing your Keras model, whether it’s the architecture, the training process, the layers or its internals, is becoming increasingly important...
What do ConvNets see? Visualizing filters with Activation Maximization
Training a ConvNet can be equal to training a black box: you start the training process, get a model that...
Visualize layer outputs of your Keras classifier with Keract
When you train Convolutional Neural Networks, you wish to understand their performance before you apply them in the real world....
Visualizing Keras CNN attention: Grad-CAM Class Activation Maps
When training image classifiers, you wish to know that it generates predictions based on what you want the model to...
Visualizing Keras CNN attention: Saliency maps
Suppose that you’re training an image classifier. You don’t have much training data. The classifier is binary and it allows...
Visualizing Keras model inputs with Activation Maximization
Deep neural networks were black boxes traditionally. The mantra “you feed them data, you’ll get a working model, but you...
