No abstract

To better explore the encapsulation efficiency of citral microcapsules through spray‐drying, amphiphilic methylcellulose (MC) was used as the emulsifier and the interior wall material, with chitosan (CTS) and alginate (ALG) as the composite external wall materials. The sequence and proportion of wall materials were compared and optimized based on the stability of the emulsions and citral content in the microcapsules. MC/CTS/ALG was the best wall material for citral microencapsulation. Formulation comprising 1.5 ml citral, 0.8 g MC, 1.0 g CTS, and 6.0 g ALG had the best entrapment of citral, resulting in citral unit content of 46.4 mg/g and the encapsulation efficiency of 82.2%. Scanning electron microscope and transmission electron microscopy images expressed distinct spherical core–shell structure of the microcapsules. Citral microcapsules absorbed water during storage, which resulted in particle size increase and cracks in the microcapsules surface. Water absorption properties in common used environments (open in air, sealed plastic bag, glass bottle) were investigated. CoCl2 microcapsules were used to intuitively express the water absorption through the color change. The results are referential for industry to select proper wall materials and storage conditions for spice microcapsules to increase the shelf life of relevant foods.

Recently, the attention mechanism combining spatial and channel information has been widely used in various deep convolutional neural networks (CNNs), proving its great potential in improving model performance. However, this usually uses 2D global pooling operations to compress spatial information or scaling methods to reduce the computational overhead in channel attention. These methods will result in severe information loss. Therefore, we propose a Spatial channel attention mechanism that captures cross-dimensional interaction, which does not involve dimensionality reduction and brings significant performance improvement with negligible computational overhead. The proposed attention mechanism can be seamlessly integrated into any convolutional neural network since it is a lightweight general module. Our method achieves a performance improvement of 2.08% on ResNet and 1.02% on MobileNetV2 in top-one error rate on the ImageNet dataset.