In recent years, convolutional neural networks (CNNs) have been widely used in various computer visual recognition tasks and have achieved good results compared with traditional methods. Image classification is one of the basic and important tasks of visual recognition, and the CNN architecture applied to other visual recognition tasks (such as object detection, object localization, and semantic segmentation) is generally derived from the network architecture in image classification. We first summarize the development history of CNNs and then analyze the architecture of various deep CNNs in image classification. Furthermore, not only the innovation of the network architecture is beneficial to the results of image classification, but also the improvement of the network optimization method or training method has improved the result of image classification. We also analyze each of these methods' effect. The experimental results of various methods are compared. Finally, the development of future CNNs is prospected.

Porous ternary metal sulfide integrated electrode materials with abundant electroactive sites and redox reactions are very promising for supercapacitors. Herein, a porous zinc cobalt sulfide nanosheet array on Ni foam (Zn-Co-S/NF) was constructed by facile growth of 2D bimetallic zinc/cobalt-based metal-organic framework (Zn/Co-MOF) nanosheets with leaf-like morphology on NF, followed by additional sulfurization. The Zn-Co-S/NF nanosheet array acted directly as a supercapacitor electrode showing much better electrochemical performance (2354.3 F g and 88.6 % retention over 1000 cycles) when compared with zinc cobalt sulfide powder (355.3 F g and 75.8 % retention over 1000 cycles), which originates from good electrical conductivity and mechanical stability, abundant electroactive sites, and facilitated transportation of electrons and electrolyte ions due to the unique nanosheet array structure. An asymmetric supercapacitor (ASC) device assembled from Zn-Co-S/NF and activated carbon electrodes can deliver a highest energy density of 31.9 Wh kg and a maximum power density of 8.5 kW kg . Most importantly, this ASC also shows good cycling stability (71.0 % retention over 10000 cycles). Furthermore, a red LED can be powered by two connected ASCs, and thus as-synthesized Zn-Co-S/NF has great potential for practical applications.

High energy density is achieved on an electrode based on ultrathin Ni-MOF nanosheet arrays grown on PANI decorated NF.

In order to enhance dielectric properties of polymer-derived SiC ceramics, a novel single-source-precursor was synthesized by the reaction of an allylhydrido polycarbosilane (AHPCS) and divinyl benzene (DVB) to form carbon-rich SiC. As expected, the free carbon contents of resultant SiC ceramics annealed at 1600 °C are significantly enhanced from 6.62 wt% to 44.67 wt%. After annealing at 900–1600 °C, the obtained carbon-rich SiC ceramics undergo phase separation from amorphous to crystalline feature where superfine SiC nanocrystals and turbostratic carbon networks are dispersed in an amorphous SiC(O) matrix. The dielectric properties and electromagnetic (EM) absorption performance of as-synthesized carbon-rich SiC ceramics are significantly improved by increasing the structural order and content of free carbon. For the 1600 °C ceramics mixed with paraffin wax, the minimum reflection coefficient (RCmin) reaches –56.8 dB at 15.2 GHz with the thickness of 1.51 mm and a relatively broad effective bandwidth (the bandwidth of RC values lower than –10 dB) of 4.43 GHz, indicating the excellent EM absorption performance. The carbon-rich SiC ceramics have to be considered as harsh environmental EM absorbers with excellent chemical stability, high temperature, and oxidation and corrosion resistance.

Leaf-like Co3O4@NiCo2O4 nanoarrays with an excellent energy storage performance was designed by a MOF-guided strategy.