基于知识蒸馏的道路交通标志识别神经网络OA北大核心CSTPCD
Lightweight Road Traffic Sign Identification Neural Network Based on Knowledge Distillation
自然场景下的交通标志识别易受到光照、遮挡和模糊等因素的干扰,从而影响检测精度;同时现有的深度学习模型参数量多、计算复杂度较高导致模型推理时间较长.提出了一种基于知识蒸馏的神经网络架构AFE-ViT(adaptive feature extraction-vision Transformer)用于道路交通标志识别,该架构由自适应特征提取模块和轻量级ViT(vision Transformer)分类器组成,其融合了图像中局部和全局特征信息,对自然场景…查看全部>>
Recognition of traffic signs in natural scenes is susceptible to interference from factors such as lighting,occlu-sions,and blurriness,which can affect detection accuracy.Additionally,existing deep learning models have a large num-ber of parameters and high computational complexity,resulting in longer model inference times.The article proposes a neural network architecture adaptive feature extraction-vision Transformer(AFE-ViT)based on knowledge distil…查看全部>>
葛怡源;于明鑫
北京信息科技大学仪器科学与技术系,北京 100192北京信息科技大学仪器科学与技术系,北京 100192
计算机与自动化
交通标识知识蒸馏自适应特征提取
traffic sign identificationknowledge distillationadaptive feature extraction
《计算机工程与应用》 2024 (19)
110-119,10
国家自然科学基金(U21A6003)北京信息科技大学勤信英才项目(5112111145).
评论