Textbook of AI

Abstract. VGGNet, which demonstrated that stacking many small 3×3 convolutions to depths of 16 or 19 layers yields excellent image classification performance, establishing a blueprint for simple, uniform deep CNN architectures.