Visformer: The Vision-friendly Transformer

The past few years have witnessed the rapid development of applying the Transformer module to vision problems. While some researchers have demonstrated that Transformerbased models enjoy a favorable ability of fitting data, there are still growing number of evidences showing that these models suffer over-fitting especially when the training data is limited. This paper offers an empirical study by performing step-bystep operations to gradually transit a Transformer-based model to a convolution-based model. The results we obtain during the transition process deliver useful messages for improving visual recognition. Based on these observations, we propose a new architecture named Visformer, which is abbreviated from the 'Vision-friendly Transformer'. With the same computational complexity, Visformer outperforms both the Transformer-based and convolution-based models in terms of ImageNet classification and object detection performance, and the advantage becomes more significant when the model complexity is lower or the training set is smaller. The code is available at https://github. com/danczs/Visformer .