Convolutional neural networks (CNN) are extensively used in image classification and object recognition applications. Therefore, we consider them in this chapter. We introduce mathematical definition of the convolution operation and its implementation by a single neuron first. Afterward, we introduce the convolution definition under Keras. Then, we explore how to form a complete CNN model under Keras. Here, we focus on feature extraction and classification blocks forming the CNN model. Afterward, we consider training and testing steps of the formed model. We then explore transfer learning to benefit from existing CNN models to be modified for our own problem. We next consider implementation steps for embedding the trained and tested CNN model on the STM32 microcontroller. Finally, we provide examples on the usage of CNN models to solve real-life problems.