Article Details

Abstract

Fine-grained flower image classification in complex natural scenarios is challenged by cluttered backgrounds, inconsistent lighting, and occlusions, which hinder the extraction of discriminative features and limit recognition accuracy. To tackle this, this study proposes an enhanced classification framework built on SE-ResNet34. By integrating the Squeeze-and-Excitation (SE) channel attention module into ResNet34's core residual blocks, the module enables the network to autonomously recalibrate feature weights. Tested and validated on the public Oxford Flowers 102 dataset, our model achieves a Top-1 accuracy of 98.30%. It not only outperforms the baseline ResNet34 by a solid 5.15 percentage points in precision but also accelerates convergence by approximately 30%. Even under challenging conditions such as occlusion and low-light scenarios, the model maintains a competitive F1-score of 97.92%.

Keywords

flower recognition; fine-grained image classification; attention mechanism; ResNet34; transfer learning

References

1. [1] Li Chao Analysis of the Importance of Flowers in Landscape Greening [J]. Real Estate Guide, 2014 (17): 221
2. [2] Li Xiangxia, Ji Xiaohui, Li Bin Deep Learning Methods for Fine grained Image Classification [J]. Computer Science and Exploration, 2021, 15 (10): 1830-1842
3. [3] Deng J，Dong W，Socher R，et al. ImageNet：A Large-Scale Hierarchical Image Database [C]//2009 IEEE Conference on Computer Vision and Pattern Recognition，2009：248-255.
4. [4] He K，Zhang X，Ren S，et al. Deep Residual Learning for Image Recognition [C]//Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition，2016：770-778.
5. [5] Simonyan K，Zisserman A. Very Deep Convolutional Networks for Large-Scale Image Recognition [J/OL].arXiv:1409.1556 [cs.CV].https://doi.org/10.48550/arXiv.1409.1556.
6. [6] Hu J，Shen L，Samuel A，et al. Squeeze-and-Excitation Networks [J]. IEEE Transactions on Pattern Analysis and Machine Intelligence，2019，42（8）：2011-2023.
7. [7] Ren Huan, Wang Xuguang A Review of Attention Mechanisms [J]. Computer Applications, 2021, 41 (S1): 1-6
8. [8] Yang Suorong, Yang Hongchao, Shen Fufu, etc Overview of Image Data Enhancement for Deep Learning [J]. Journal of Software, 2025, 36 (3): 1390-1412
9. [9] Nilsback M E，Zisserman A. Automated Flower Classification Over a Large Number of Classes [C]//2008 Sixth Indian Conference on Computer Vision，Graphics & Image Processing，2008：722-729.
10. [10] Pan S J，Yang Q. A Survey on Transfer Learning [J].IEEE Transactions on Knowledge and Data Engineering，2009，22（10）：1345-1359.
11. [11] Gao Shidi, Yang Mingyue, Jiang Dan, etc Research on Flower Variety Identification Based on Attention Mechanism and Transfer Learning [J]. Software, 2022, 43 (10): 46-50
12. [12] Zhang Mengyu Flower recognition based on ResNet and attention mechanism [J]. Computer and Modernization, 2021 (4): 61-67

Funding

National Natural Science Foundation of China (32460440)

About the authors

Yang Jiaxin (2005-), male, Han ethnicity, from Qingyang, Gansu Province, is currently pursuing a bachelor's degree. His research interests include computer vision and fine-grained image classification; Wu Xuemei (2006-), female, Han ethnicity, from Tianshui, Gansu Province, undergraduate studies, research direction: deep learning; Cui Wenxin (2006-), male, Han ethnicity, from Baoding, Hebei Province, undergraduate studies, research direction: computer vision; He Jianchen (2005-), female, Han ethnicity, from Wuwei, Gansu Province, undergraduate student, research direction: data mining; Corresponding author: Li Yongbiao (1996-), male, Han ethnicity, from Zhangye, Gansu Province, teaching assistant, master's degree, research direction: artificial intelligence.