We propose a new neural network design paradigm Reversible Column Network (RevCol). The main body of RevCol is composed of multiple copies of subnetworks, named columns respectively, between which multi-level reversible connections are employed. Such architectural scheme attributes RevCol very different behavior from conventional networks: during forward propagation, features in RevCol are learned to be gradually disentangled when passing through each column, whose total information is maintained rather than compressed or discarded as other network does. Our experiments suggest that CNN-style RevCol models can achieve very competitive performances on multiple computer vision tasks such as image classification, object detection and semantic segmentation, especially with large parameter budget and large dataset. For example, after ImageNet-22K pre-training, RevCol-XL obtains 88.2% ImageNet-1K accuracy. Given more pre-training data, our largest model RevCol-H reaches 90.0% on ImageNet-1K, 63.8% APbox on COCO detection minival set, 61.0% mIoU on ADE20k segmentation. To our knowledge, it is the best COCO detection and ADE20k segmentation result among pure (static) CNN models. Moreover, as a general macro architecture fashion, RevCol can also be introduced into transformers or other neural networks, which is demonstrated to improve the performances in both computer vision and NLP tasks. We release code and models at https://github.com/megvii-research/RevCol

翻译：我们提出了一种新的神经网络设计范式——可逆列网络（RevCol）。RevCol的主体由多个子网络副本构成（分别称为列），这些列之间采用了多级可逆连接。这种架构设计使RevCol展现出与传统网络截然不同的行为：在前向传播过程中，RevCol中的特征在通过每一列时会逐渐被学习解耦，其总信息量得以保持，而非像其他网络那样被压缩或丢弃。我们的实验表明，基于CNN的RevCol模型在图像分类、目标检测和语义分割等多个计算机视觉任务上能够取得极具竞争力的性能，特别是在大参数量和大规模数据集场景下。例如，经过ImageNet-22K预训练后，RevCol-XL在ImageNet-1K上获得了88.2%的准确率。在更多预训练数据的支撑下，我们最大的模型RevCol-H在ImageNet-1K上达到90.0%，在COCO检测minival集上达到63.8%的APbox，在ADE20k分割上达到61.0%的mIoU。据我们所知，这是纯（静态）CNN模型中最佳的COCO检测和ADE20k分割结果。此外，作为一种通用的宏观架构范式，RevCol还可引入Transformer或其他神经网络中，并被证明能提升计算机视觉和自然语言处理任务的性能。我们在https://github.com/megvii-research/RevCol 发布了代码和模型。