Fashion clothes classification encompasses spotting and identifying items of clothing in an image. This area of research has involved using deep neural networks to make an impact in the field of social media, e-commerce and fashion world. In this paper, we propose an attention-driven technique for tackling visual fashion clothes analysis in images, aiming to achieve clothing category classification and attribute prediction by producing regularised landmark layouts. For enhancing clothing classification, our fashion model incorporates two attention pipelines: landmark-driven attention and spatial–channel attention. These attention pipelines allow our model to represent multiscale contextual information of landmarks, thus improving the efficiency of classification by identifying the important features and locating where they exist in an input image. We evaluated the proposed network on two large-scale benchmark datasets: DeepFashion-C and fashion landmark detection (FLD). Experimental results show that the proposed architecture involving deep neural network outperforms other recently reported state-of-the-art techniques in the classification of fashion clothes.

中文翻译：

---

一种用于时装分类的改进的地标驱动和空间通道注意卷积神经网络

时装分类包括在图像中发现和识别服装项目。该研究领域涉及使用深度神经网络对社交媒体、电子商务和时尚界产生影响。在本文中，我们提出了一种注意力驱动技术，用于处理图像中的视觉时尚服装分析，旨在通过生成正则化的地标布局来实现服装类别分类和属性预测。为了增强服装分类，我们的时装模型结合了两个注意力管道：地标驱动注意力和空间通道注意力。这些注意力管道使我们的模型能够表示地标的多尺度上下文信息，从而通过识别重要特征并定位它们在输入图像中的位置来提高分类效率。我们在两个大规模基准数据集上评估了提议的网络：DeepFashion-C 和时尚地标检测 (FLD)。实验结果表明，所提出的涉及深度神经网络的架构在时尚服装分类方面优于其他最近报道的最先进技术。