Window attention, position embeddings, and high resolution finetuning are
core concepts in the modern transformer era of computer vision. However, we
find that naively combining these near ubiquitous components can have a
detrimental effect on performance. The issue is simple: interpolating position
embeddings while using window attention is wrong. We study two state-of-the-art
methods that have these three components, namely Hiera and ViTDet, and find
that both do indeed suffer from this bug. To fix it, we introduce a simple
absolute window position embedding strategy, which solves the bug outright in
Hiera and allows us to increase both speed and performance of the model in
ViTDet. We finally combine the two to obtain HieraDet, which achieves 61.7 box
mAP on COCO, making it state-of-the-art for models that only use ImageNet-1k
pretraining. This all stems from what is essentially a 3 line bug fix, which we
name "absolute win".

在现代计算机视觉转换器时代，窗口注意力、位置嵌入和高分辨率微调是核心概念。然而，我们发现，天真地结合这些几乎无处不在的组件可能会对性能产生不利影响。我们研究了两种具有这三个组件的最先进方法，即 Hiera 和 ViTDet，发现两者都确实遭受了这个错误的困扰。为了解决这个问题，我们引入了一种简单的绝对窗口位置嵌入策略，在 Hiera 中彻底解决了该错误，并允许我们在 ViTDet 中提高模型的速度和性能。最后，我们将两者结合起来得到 HieraDet，在 COCO 上实现了 61.7 的盒子 mAP，成为仅使用 ImageNet-1k 预训练的模型的最先进技术。这一切都源自本质上是一个三行错误修复，我们将其命名为 “绝对获胜”。