基于改进ConvNeXt的皮肤镜图像分类方法

皮肤癌是最致命的癌症之一,对皮肤镜图像进行精确分类尤为关键,然而现有的皮肤镜图像存在形态复杂、样本数量较少的问题,导致现有的自动分类方法难以提取图像特征信息,误判率较高。提出一种改进ConvNeXt的方法,并构建SE-SimAM-ConvNeXt模型。以ConvNeXt为基础网络,加入SimAM无参注意力模块,提升网络的特征提取能力,并在基础网络中引入通道注意力机制,增强ConvNeXt对潜在关键特征的挖掘能力。在训练初始时加入预热机制Cosine Warmup,在该过程中使用余弦函数值进行学习率的衰减,进一步加速ConvNeXt的收敛,提高ConvNeXt模型的分类能力。在HAM10000皮肤数据集上的实验结果表明,该模型的分类准确率、精确度、召回率、特异性分别为92.9%、85.3%、78.0%、97.5%,具有较好的皮肤镜图像分类能力,对皮肤癌病变的辅助诊断有一定程度的应用价值,可帮助皮肤科医生对皮肤癌做进一步的诊断。

Application of the Plate Crown Prediction

The program of the plate crown optimal prediction applied in the plate mill of Wuhan Iron and Steel(Group) Co. Ltd. is described. Optimal prediction is programmed in ALPHA 4000. The original profile, wear, expansion and deformation of work rolls are introduced into the program. The plate crown is controlled by the original profile of work rolls and the draft whose objective is the optimal plate crown. The run schedule is also optimized through the optimal prediction of the plate crown.

Thermal-mechanical properties of short carbon fiber reinforced geopolymer matrix composites subjected to thermal load

Short carbon fiber preform reinforced geopolymer composites containing different contents of α-Al2O3 filler （Cr（a-Al2O3）/geopolymer composites） were fabricated, and the effects of heat treatment temperatures up to 1 200 ℃ on the thermal-mechanical properties were studied. The results show that the thermal shrinkage in the direction perpendicular to the lamination of the composites gradually increases with the increase of the heat treatment temperatures from room temperature （25 ℃ ） to 1000 ℃. However, the composites in the direction parallel to the lamination show an expansion behavior. Beyond 1 000℃, in the two directions the composites exhibit a larger degree of shrinkage due to the densification and crystallization. The mechanical properties of the composites show the minimum values in the temperature range from 600 to 800 ℃ as the hydration water of geopolymer matrix is lost. The addition of α-Al2O3 particle filler into the composites clearly increases the onset crystalline temperature of leucite （KAlSi2O6） from the amorphous geopolymer matrix. In addition, the addition of α-Al2O3 particles into the composites can not only help to keep volume stable at high temperatures but also effectively improve the mechanical properties of the composites subjected to thermal load to a certain extent. The main toughening mechanisms of the composites subjected to thermal load are attributed to fiber pulling-out.