Image steganalysis aims to detect whether secret information is hidden in an image. This technique has critical applications in the field of information security. Most existing methods combine popular computer vision components for design without profoundly exploring the key factors applicable to image steganalysis. This letter reveals the limitations of existing feature passing and downsampling methods for image steganalysis tasks. We found that existing methods that pass shallow features through residual connections cannot cope with the problem of feature disappearance during network forward. In addition, the information-reducing downsampling methods used by these methods suppress the expression of steganographic features. To address these issues, we propose a feature enhancement passing module (FEPM) to help pass shallow features to deep layers and an attention downsampling module (ADM) to perform attention learning on full-resolution features. Combining these two structures, we designed an ultra-lightweight and highprecision image steganalysis network, FPNet, which contains only 0.16M parameters. The results of several experiments in the same environment show that our method outperforms current state-of-the-art methods in several aspects, including detection accuracy and computational effort. The code is available at https://github.com/henryccl/FPNet .

The accurate and automatic segmentation of retinal vessels from fundus images is critical for the early diagnosis and prevention of many eye diseases, such as diabetic retinopathy (DR). Existing retinal vessel segmentation approaches based on convolutional neural networks (CNNs) have achieved remarkable effectiveness. Here, we extend a retinal vessel segmentation model with low complexity and high performance based on U-Net, which is one of the most popular architectures. In view of the excellent work of depth-wise separable convolution, we introduce it to replace the standard convolutional layer. The complexity of the proposed model is reduced by decreasing the number of parameters and calculations required for the model. To ensure performance while lowering redundant parameters, we integrate the pre-trained MobileNet V2 into the encoder. Then, a feature fusion residual module (FFRM) is designed to facilitate complementary strengths by enhancing the effective fusion between adjacent levels, which alleviates extraneous clutter introduced by direct fusion. Finally, we provide detailed comparisons between the proposed SepFE and U-Net in three retinal image mainstream datasets (DRIVE, STARE, and CHASEDB1). The results show that the number of SepFE parameters is only 3% of U-Net, the Flops are only 8% of U-Net, and better segmentation performance is obtained. The superiority of SepFE is further demonstrated through comparisons with other advanced methods.