Resnet in Resnet: Generalizing Residual Architectures
This work addresses the need for more effective neural network architectures in computer vision, offering incremental improvements over existing ResNets.
The authors tackled the problem of improving residual networks (ResNets) for computer vision tasks by introducing Resnet in Resnet (RiR), a dual-stream architecture that generalizes ResNets and CNNs with no computational overhead, resulting in consistent performance gains over ResNets, outperforming similar architectures on CIFAR-10, and achieving a new state-of-the-art on CIFAR-100.
Residual networks (ResNets) have recently achieved state-of-the-art on challenging computer vision tasks. We introduce Resnet in Resnet (RiR): a deep dual-stream architecture that generalizes ResNets and standard CNNs and is easily implemented with no computational overhead. RiR consistently improves performance over ResNets, outperforms architectures with similar amounts of augmentation on CIFAR-10, and establishes a new state-of-the-art on CIFAR-100.