Recent advances in convolutional neural networks (CNNs) for image feature extraction have achieved extraordinary performance, but back-propagation algorithms tend to fall into local minima. To alleviate this problem, this paper proposes a coordinate attention-support vector machine-convolutional neural network (CASVM). This proposed to enhance the model’s ability by introducing coordinate attention while obtaining enhanced image features. Training is carried out by back-propagating the loss function of support vector machines (SVMs) to improve the generalization capability, which can effectively avoid falling into local optima. The image datasets used in this study for benchmark experiments are Fashion-MNIST, Cifar10, Cifar100, and Animal10. Experimental results show that compared with softmax, CASVM can improve the image classification accuracy of the original model under different image resolution datasets. Under the same structure, CASVM shows better performance and robustness and has higher accuracy. Under the same network parameters, the loss function of CASVM enables the model to realize a lower loss value. Among the standard CNN models, the highest accuracy rate can reach 99%, and the optimal number of accuracy indicators is 5.5 times that of softmax, whose accuracy rate can be improved by up to 56%.
Recent advancements in transformer-based monocular 3D object detection techniques have exhibited exceptional performance in inferring 3D attributes from single 2D images. However, most existing methods rely on resource-intensive transformer architectures, which often lead to significant drops in computational efficiency and performance when handling long sequence data. To address these challenges and advance monocular 3D object detection technology, we propose an innovative network architecture, MonoMM, a Multi-scale \textbf{M}amba-Enhanced network for real-time Monocular 3D object detection. This well-designed architecture primarily includes the following two core modules: Focused Multi-Scale Fusion (FMF) Module, which focuses on effectively preserving and fusing image information from different scales with lower computational resource consumption. By precisely regulating the information flow, the FMF module enhances the model adaptability and robustness to scale variations while maintaining image details. Depth-Aware Feature Enhancement Mamba (DMB) Module: It utilizes the fused features from image characteristics as input and employs a novel adaptive strategy to globally integrate depth information and visual information. This depth fusion strategy not only improves the accuracy of depth estimation but also enhances the model performance under different viewing angles and environmental conditions. Moreover, the modular design of MonoMM provides high flexibility and scalability, facilitating adjustments and optimizations according to specific application needs. Extensive experiments conducted on the KITTI dataset show that our method outperforms previous monocular methods and achieves real-time detection.
25 October 2025 in Pattern Analysis and Applications