we’re going to talk about Attention and Transformers. This is a really exciting and important topic. These are ideas that have truly come to dominate not just computer vision, but really all of modern deep learning over the last several years. So this will be a really foundational for understanding the state of the art.
How do we build CNNs? This isn’t just about stacking layers arbitrarily. It involves a nuanced understanding of the various types of layers in CNNs: convolution, pooling, fully connected, and others. Once we’ve designed an architecture, the training process is, of course, where the learning occurs. It’s where the network adapts its parameters to minimize the loss function on the training data.
Regularization and optimization are the two absolute critical concepts in deep learning that allow our model to learn effectively and generalize well to new data
From early image processing to the rise of neural networks, now we will look back at the evolution of computer vision and deep learning.