To solve the problem of misclassification of traffic signs under various scale conditions, similar categories with insufficient illumination, image blurriness, partial occlusions, etc., a light-weight multi-scale attention convolutional neural network is introduced. Use a low-resolution image for edge extraction combined with color texture; add residual block layers at last step to increase depth of semantic information reconstruction. Multi-scale convolutional Branches are used to enhance the model's capability of recognizing distant Small Targets and scales changing sample. Introduce a Channel-pace Attention Module to enhance target detection and Tracking Performance in Key Regions. Focal Loss, label smoothing algorithms to improve the model at test time. The top-1 accuracy of experiment results from the German Traffic Sign Recognition Benchmark (GTSRB) was as high as 99.08%, and its F1 scores were around 98.97%. It has strong robustness under conditions such as poor lighting, raindrops, haze, occlusion, and faraway small targets, with moderate model size, computing power consumption, and inference speed. It is appropriate to apply this design in practice.
Zhihao Zou· International Conference on...· 0 citations
ViTs have achieved excellent performance in image recognition tasks, but their large parameter counts and high computational complexity limit their deployment on resource-constrained devices. Most existing ViT pruning methods adopt class-agnostic pruning strategies, which fail to distinguish the diverse structural requirements of different target classes. As a result, they are prone to removing critical features, leading to class-wise accuracy imbalance in practical deployment. To address this issue, this paper proposes a class-aware joint pruning framework for ViTs, which collaboratively compresses the model from two orthogonal dimensions: MLP neurons and visual tokens. Specifically, (1) based on first-order Taylor expansion, we quantify the contribution of each MLP neuron to the target classes and adaptively prune redundant neurons to achieve structured compression, followed by lightweight fine-tuning on the target class subset; (2) we propose a Class-Guided Token Selection (CGTS) method, which constructs class prototype vectors using a few support samples of the target classes and then dynamically selects patch tokens that are semantically highly relevant to the target classes during inference in a zero-shot manner, requiring no additional training or fine-tuning. The two modules complement each other, achieving dual compression from the parameter dimension and the inference data dimension. Experiments on CIFAR-100 and TinyImageNet datasets using DeiT-Tiny/Small models demonstrate that, compared with state-of-the-art pruning methods, our method reduces GMACs on target class subsets by up to 48%, improves inference speed by nearly 50%, and requires only 0.8 KB of additional storage overhead per subset, ultimately achieving a superior trade-off among accuracy, computational efficiency, and storage overhead.
Zhi-Yue Li, Tong Liu, Feng Huang et al.· Big Data and Cognitive Compu...· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
