Co-Scale Conv-Attentional Image Transformers
This work addresses image classification and vision tasks for computer vision researchers, presenting an incremental improvement by combining existing mechanisms in a novel way.
The paper tackled the problem of improving image classification by introducing CoaT, a Transformer-based model with co-scale and conv-attentional mechanisms, achieving superior classification results on ImageNet compared to similar-sized CNNs and Transformers, and demonstrating effectiveness in downstream tasks like object detection and instance segmentation.
In this paper, we present Co-scale conv-attentional image Transformers (CoaT), a Transformer-based image classifier equipped with co-scale and conv-attentional mechanisms. First, the co-scale mechanism maintains the integrity of Transformers' encoder branches at individual scales, while allowing representations learned at different scales to effectively communicate with each other; we design a series of serial and parallel blocks to realize the co-scale mechanism. Second, we devise a conv-attentional mechanism by realizing a relative position embedding formulation in the factorized attention module with an efficient convolution-like implementation. CoaT empowers image Transformers with enriched multi-scale and contextual modeling capabilities. On ImageNet, relatively small CoaT models attain superior classification results compared with similar-sized convolutional neural networks and image/vision Transformers. The effectiveness of CoaT's backbone is also illustrated on object detection and instance segmentation, demonstrating its applicability to downstream computer vision tasks.