The paper presents SA2-Net, a scale-aware attention network for microscopic image segmentation. Microscopic image segmentation, assigning semantic labels to each pixel in a microscopic image, is a challenging task. Existing frameworks, often based on convolutional neural networks (CNNs), struggle to capture long-range dependencies. The proposed SA2-Net uses attention-guided method and multi-scale feature learning to handle diverse structures within microscopic images. It also introduces a novel up-sampling strategy called the Adaptive Up-Attention (AuA) module for improved localization of microscopic regions. The effectiveness of SA2-Net is demonstrated through experiments on five challenging datasets.
Publication date: 28 Sep 2023
Project Page: https://github.com/mustansarfiaz/SA2-Net
Paper: https://arxiv.org/pdf/2309.16661
