Fine-grained Classification Model of High-frequency Components Based on Reinforced Spatial Transformation
LI Guo-peng, LUO Jian-qiao, ZENG Bao-zhi, XIONG Ying, LI Bai-lin
Fine-grained Classification Model of High-frequency Components Based on Reinforced Spatial Transformation
表1 数据集划分 |
训练集 | 验证集 | 测试集 | |
---|---|---|---|
图像数量 | 5434 | 1811 | 1811 |
表2 不同方法的高频元件分类性能表 |
方法 | 主干网络 | 准确率 | 推理时间/(ms/张) |
---|---|---|---|
Base | Moblilenet_V3 | 0.8040 | 32.04 |
MC | Moblilenet_V3 | 0.8353 | 30.02 |
STB | VGG16 | 0.8601 | 31.65 |
JSST | ResNet34 | 0.8489 | 30.40 |
DFLA | ResNet50 | 0.8563 | 30.84 |
STN | Moblilenet_V3 | 0.8857 | 31.73 |
rSTN-S | Moblilenet_V3 | 0.9029 | 33.94 |
rSTN-L | Moblilenet_V3 | 0.8982 | 32.11 |
rSTN | Moblilenet_V3 | 0.9212 | 34.23 |
表3 加权梯度类激活映射(Grad-CAM)实验结果 |
类别 | 样本 | Base | MC | rSTN | |||
---|---|---|---|---|---|---|---|
网络输入 | 类别激活图 | 网络输入 | 类别激活图 | 网络输入 | 类别激活图 | ||
G1 | S1 | ||||||
S2 | |||||||
G2 | S3 | ||||||
S4 | |||||||
G3 | S5 | ||||||
S6 |
表4 网络预测样本真实类别概率TOP5表 |
样本 | Base | MC | rSTN |
---|---|---|---|
S1 | 0.852 | 0.886 | 0.964 |
S2 | 0.849 | 0.879 | 0.955 |
S3 | 0.853 | 0.881 | 0.944 |
S4 | 0.848 | 0.875 | 0.956 |
S5 | 0.858 | 0.880 | 0.952 |
S6 | 0.844 | 0.875 | 0.948 |
1 |
魏涛,张丹,左敦稳,等.面向薄壁多腔类结构件加工特征识别方法[J].计算机集成制造系统,2017,23(12):2683-2691.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
2 |
李国瑞,何小海,吴晓红,等.基于语义信息跨层特征融合的细粒度鸟类识别[J].计算机应用与软件,2020,37(4):132-136.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
3 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
4 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
5 |
顾攀,张烽栋.基于神经网络的图像弱监督语义分割算法[J].计算机应用与软件,2018,35(2):284-288.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
6 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
7 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
8 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
9 |
冀中,赵可心,张锁平,等.基于空间变换双线性网络的细粒度鱼类图像分类[J].天津大学学报:自然科学与工程技术版.2019,52(5):475-82.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
10 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
11 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
12 |
陈绪,陈志澜.基于迁移学习的零件识别方法研究[J].制造业自动化,2019,41(8):81-86.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
13 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
14 |
蒲云洁,王学渊.基于CUDA加速的多模态膝关节图像配准[J].制造业自动化,2022,44(3):11-14+26.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
15 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
16 |
祝礼佳,刘桂华,林杰.sss基于卷积神经网络的磁瓦表面缺陷识别[J].制造业自动化,2022,44(3):48-53.
{{custom_citation.content}}
{{custom_citation.annotation}}
|
17 |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
{{custom_ref.label}} |
{{custom_citation.content}}
{{custom_citation.annotation}}
|
/
〈 | 〉 |