Residual Block

A building block that eases the training of deep CNNs by carrying information directly through identity connections.

Residual blocks are one of the most important architectural innovations that made very deep networks trainable. Instead of forcing layers to learn full transformations directly, they can learn residual transformations added to the input, often yielding more stable optimization. This improves gradient flow and makes deeper CNNs practical. A large portion of modern vision architectures has been shaped by residual design principles.