Convolution exploits locality for efficiency at a cost of missing long range context. Self-attention has been adopted to augment CNNs with non-local interactions. Recent works prove it possible to stack self-attention layers to obtain a fully attentional network by restricting the attention to a local region. In this paper, we attempt to remove this constraint by factorizing 2D self-attention into two 1D selfattentions. This reduces computation complexity and allows performing attention within a larger or even global region. In companion, we also propose a position-sensitive self-attention design. Combining both yields our position-sensitive axial-attention layer, a novel building block that one could stack to form axial-attention models for image classification and dense prediction. We demonstrate the effectiveness of our model on four large-scale datasets. In particular, our model outperforms all existing stand-alone self-attention models on ImageNet. Our Axial-DeepLab improves 2.8% PQ over bottom-up state-of-the-art on COCO test-dev. This previous state-of-the-art is attained by our small variant that is 3.8x parameter-efficient and 27x computation-efficient. Axial-DeepLab also achieves state-of-the-art results on Mapillary Vistas and Cityscapes.

No abstract

Objectives-Contemporary methods of dentin bonding could create hybrid layers (HLs) containing voids and exposed, demineralized collagen fibers. Proanthocyanidins (PA) have been shown to crosslink and strengthen demineralized dentin collagen, but their effects on collagen degradation within the HL have not been widely studied. The purpose of this study was to compare the morphological differences of HLs created by BisGMA/HEMA model adhesives with and without the addition of grape seed extract PA under conditions of enzymatic collagen degradation.Methods-Model adhesives formulated with and without 5% PA were bonded to the acid etched dentin. Five-μm-thick sections cut from the bonded specimens were stained with Goldner's trichrome. The specimens were then exposed to 0.1% collagenase solution for zero, one, or six days. Following collagenase treatment, the specimens were analyzed with SEM/TEM.Results-Staining did not reveal a difference in the HLs created with the two adhesives. SEM showed the presence of intact collagen fibrils in all collagenase treatment conditions for specimens bonded with adhesive containing PA. These integral collagen fibrils were not observed in the specimens bonded with adhesive without PA after the same collagenase treatment. TEM confirmed that the specimens containing PA still showed normal collagen fibril organization and dimensions after treatment with collagenase solution. In contrast, disorganized collagen fibrils in the interfacial zone lacked the typical cross-banding of normal collagen after collagenase treatment for specimens without PA.Conclusions-The presence of grape seed extract PA in dental adhesives may inhibit the biodegradation of unprotected collagen fibrils within the HL.

No abstract