Involution: Inverting the Inherence of Convolution for Visual Recognition
This work addresses a foundational problem in deep learning for vision tasks by introducing a new operator that could replace convolution, offering significant efficiency and accuracy gains.
The paper tackles the limitations of standard convolution in visual recognition by proposing a novel atomic operation called involution, which inverts convolution's spatial-agnostic and channel-specific principles, resulting in improved performance on benchmarks like ImageNet (up to 1.6% top-1 accuracy) and reduced computational costs (down to 57%).
Convolution has been the core ingredient of modern neural networks, triggering the surge of deep learning in vision. In this work, we rethink the inherent principles of standard convolution for vision tasks, specifically spatial-agnostic and channel-specific. Instead, we present a novel atomic operation for deep neural networks by inverting the aforementioned design principles of convolution, coined as involution. We additionally demystify the recent popular self-attention operator and subsume it into our involution family as an over-complicated instantiation. The proposed involution operator could be leveraged as fundamental bricks to build the new generation of neural networks for visual recognition, powering different deep learning models on several prevalent benchmarks, including ImageNet classification, COCO detection and segmentation, together with Cityscapes segmentation. Our involution-based models improve the performance of convolutional baselines using ResNet-50 by up to 1.6% top-1 accuracy, 2.5% and 2.4% bounding box AP, and 4.7% mean IoU absolutely while compressing the computational cost to 66%, 65%, 72%, and 57% on the above benchmarks, respectively. Code and pre-trained models for all the tasks are available at https://github.com/d-li14/involution.