Dual-Branch Chronological Clustering Network for Bronze Inscriptions with Improved Multiscale CBAM

Jingwen Ding; Ying Lu; Huiqin Wang; Ke Wang; Zhan Wang

doi:10.3788/LOP242364

Laser & Optoelectronics Progress, Volume. 62, Issue 10, 1037003(2025)

Dual-Branch Chronological Clustering Network for Bronze Inscriptions with Improved Multiscale CBAM

Jingwen Ding1... Ying Lu1, Huiqin Wang1,*, Ke Wang1 and Zhan Wang2 |Show fewer author(s)

¹School of Information & Control Engineering, Xi'an University of Architecture & Technology, Xi'an 710055, Shaanxi , China

²Shaanxi Provincial Institute of Cultural Relics Protection, Xi'an 710075, Shaanxi , China

show less

Abstract Get PDF(in Chinese)

Figures & Tables(18)

Fig. 1. Overall framework diagram of bronze inscription age clustering network

Download full size

Fig. 2. CBAM structure diagram

Download full size

Fig. 3. Multiscale channel attention module structure

Download full size

Fig. 4. Multiscale spatia attention module structure

Download full size

Fig. 5. Residual block structure

Download full size

Fig. 6. DeepCluster structure

Download full size

Fig. 7. Flowchart of the dual-branch deep clustering network for inscription chronological analysis

Download full size

Fig. 8. Comparison of images before and after preprocessing.（a） Original image；（b） denoised；（c） horizontal flip

Download full size

Fig. 9. Loss curves. (a) Dense121; (b) ResNet18; (c) ResNet50

Download full size

Fig. 10. Accuracy comparison of different algorithms

Download full size

Fig. 11. Comparison of computational complexity and precision across different algorithms

Download full size

Fig. 12. Visualization of confidence in misdated samples

Download full size

Table 1. Feature extraction network structure and parameters

View table

Table 1. Feature extraction network structure and parameters

Layer name	Input dimension	Output dimension	Parameters
Conv1	224×224×3	112×112×64	7×7×3×64
MaxPool	112×112×64	56×56×64
Conv2_x+CBAM	56×56×64	56×56×256	1×1，3×3，1×1
Conv3_x+CBAM	56×56×256	28×28×512	1×1，3×3，1×1
Conv4_x+CBAM	28×28×512	14×14×1024	1×1，3×3，1×1
Conv5_x+CBAM	14×14×1024	7×7×2048	1×1，3×3，1×1
Global AvgPool	7×7×2048	1×1×2048

Table 2. Distribution of experimental datasets
View table
Table 2. Distribution of experimental datasets
Dynasty Training set Validation set Test set
Shang 641 71 71
Western Zhou 1395 155 155
Eastern Zhou 1309 145 145

Table 3. Statistical analysis of ablation experiment results

View table

Table 3. Statistical analysis of ablation experiment results

Ablation experiment	Shang	Western Zhou	Eastern Zhou
Ablation group ①	65.79	83.82	77.93
Ablation group ②	70.29	84.41	78.92
Ablation group ③	72.31	84.66	79.02
Ablation group ④	84.09	85.81	78.62
Ablation group ⑤	92.86	86.31	89.12

Table 4. Visualization results of attention in different groups
View table
Table 4. Visualization results of attention in different groups
Dynasty Ablation group ② Ablation group ③ Ablation group ④ Ablation group ⑤
Shang
Western Zhou
Eastern Zhou

Table 5. Performance comparison of different algorithms

View table

Table 5. Performance comparison of different algorithms

Algorithm	Shang			Western Zhou			Eastern Zhou
Algorithm	precision	recall	F1	precision	recall	F1	precision	recall	F1
VGG16	68.97	56.34	62.02	78.49	94.19	85.63	77.95	68.28	72.79
AlexNet	70.42	64.79	76.03	82.39	93.55	87.61	81.38	81.38	81.38
Dense121	83.56	85.92	84.72	81.67	94.84	87.76	71.72	74.48	82.13
MobileV3	80.00	39.44	52.83	78.86	89.03	83.64	69.57	77.24	73.20
GoogleNet	80.36	63.38	70.87	80.90	92.90	86.49	81.75	77.24	79.43
Proposed algorithm	92.86	73.24	81.89	86.31	93.55	89.78	89.12	90.34	89.73

Table 6. Comparison of different clustering algorithms
View table
Table 6. Comparison of different clustering algorithms
Category Alex_Cluster^［9］ BIRCH_Cluster^［10］ Proposed algorithm
OA 86.25 79.89 89.43
Shang 73.91 66.07 92.86
Western Zhou 82.54 79.44 86.31
Eastern Zhou 82.57 72.13 89.12

Tools

Get Citation

Copy Citation Text

Jingwen Ding, Ying Lu, Huiqin Wang, Ke Wang, Zhan Wang. Dual-Branch Chronological Clustering Network for Bronze Inscriptions with Improved Multiscale CBAM[J]. Laser & Optoelectronics Progress, 2025, 62(10): 1037003

Download Citation

EndNote(RIS)BibTex Plain Text

Set citation alerts for article

Save article for my favorites