基于改进MobileFaceNet的人脸识别方法北大核心

针对目前大多数人脸识别算法参数多、计算量大,难以部署到移动端和嵌入式设备中的问题,提出了一种基于改进MobileFaceNet的人脸识别方法。通过对MobileFaceNet模型结构的调整,将bottleneck模块优化为sandglass模块,改良深度卷积和逐点卷积的相对位置,适当增大sandglass模块的输出通道数,从而减少特征压缩时的信息丢失,增强人脸空间特征的提取。实验结果表明:改进后的方法在LFW测试数据集上准确率达99.15%,模型大小和计算量分别仅为原算法的61%和45%,验证了所提方法的有效性。     

火麻仁油微乳凝胶面膜的制备及初步评价

目的:制备火麻仁油微乳凝胶补水面膜并对其理化性质、保湿性和抗氧化活性进行评价.方法:通过伪三元相图确定火麻仁油微乳最佳处方,采用单因素及正交试验优化火麻仁油微乳凝胶处方,载入保湿活性物质制得火麻仁油微乳凝胶面膜并对其质量、稳定性、保湿性和抗氧化活性进行评价.结果:火麻仁油微乳最佳处方:火麻仁油0.2％、PEG-20氢化蓖麻油0.53％、1,3丙二醇0.27％.火麻仁油微乳凝胶最佳处方为卡波姆-940含量1％,丙三醇含量5％,稀释倍数30倍.制得的微乳凝胶澄清透明、涂展性较好、易清洗、不油腻;平均粒径为(30.91±0.31)nm,PDI为0.299±0.017,电位为(-3.54±0.43)mV;黏度为494725.15 mPa·s;pH值为5.50±0.01;保湿效果优于市售某凝胶面膜;火麻仁油微乳凝胶面膜对DPPH自由基清除率高于火麻仁油毛油,质量浓度为6 mg·mL-1时可达62.39％,IC50为3.71 mg·mL-1.结论:火麻仁油微乳凝胶面膜安全、稳定,制备工艺简单,对皮肤具有良好的补水和抗氧化效果,具备开发和推广的价值.     

Surface Modification of Co3O4 Nanosheets to Promote Peroxymonosulfate Activation for Degradation of Atrazine

Catalysis Letters - This study proposed a new facile route to rational creating oxygen-vacancy (Vo)-rich surface of Co3O4 nanosheets by acetic acid leaching. The acid leached Co3O4 nanosheets was...     

GC-MS法测定缬草残渣中的化学组分

分别使用95％乙醇和四氯化碳作为提取溶剂,在78℃左右,对已提取过挥发油的缬草残渣中化学成分进行提取,借用气相色谱-质谱联用方法(GC-MS)对提取浸膏中化学组分进行分析.结果显示:以乙醇作为提取溶剂条件下,共鉴定出38中化学组分,数据出峰相对总质量分数为81.7％;以四氯化碳为提取溶剂条件下,共鉴定出44种化学组分,数据出峰相对总质量分数为84.4％.     

Mechanical properties of biomimetic ceramic with Bouligand architecture

Biomimetic Bouligand architecture is constructed in the ceramic to improve its toughness. Firstly, unidirectional carbon fiber-reinforced ZrB₂-SiC ceramic films are achieved through a vacuum-assisted filtration method using graphene oxide. Then, ceramic films are helically assembled at a fixed angle of 30° in the graphite die based on the fiber orientation. Finally, the spark plasma sintering method was utilized to densify helical assembly carbon fiber/ceramic films. By constructing Bouligand structure, high fracture toughness (7.4 MPa·m^0.5) and work of fracture (∼1055 J/m²) are achieved in ZrB₂-based ceramic. The toughening mechanisms mainly are crack deflection, twisting and branching, carbon fiber pulling out, and bridging.     

Bio-Based,Self-Crosslinkable Eucommia ulmoides Gum/Silica Hybrids with Body Temperature Triggering Shape Memory Capability

At present, the synthesis of body temperature triggering shape memory polymers usually requires elaborate structural design, which limits their wide application. Herein, starting from bio-based Eucommia ulmoides gum (EUG), a series of EUG/silica hybrids (ESHs) are prepared through a facile one-pot process, in which EUG is epoxied and then self-crosslinked with SiO₂ by epoxy ring-open reaction. Varying the amount of H₂O₂, the shape memory transition temperature (T_trans) of ESHs is adjusted to 47.4–36.6 ℃, which is close to human body temperature (37 ℃). Among them, ESH-17 exhibited the best body temperature triggering shape memory ability (T_trans = 36.6 ℃), which can restore the permanent shape within 60 s at 37 ℃ with a shape fixity ratio of 99% and shape recovery ratio near 100%. In addition, the shape memory mechanism is discussed and shows some application scenarios of ESHs. The as-produced materials can be used as smart biomaterials such as self-tightening sutures, self-sealing root canal filling materials, and so on.