Fine‑Grained Flower Image Classification Based on Neural Network Architecture Search

doi:10.15933/j.cnki.1004-3268.2024.05.018

Abstract

Abstract: To enhance the automation of deep convolutional neural network(CNN)design and improve fine‑grained flower image classification accuracy，an advanced neural network search approach based on differentiable architecture search(DARTS) was proposed.This method automatically constructed fine‑grained flower image classification models.Initially,an attention‑convolution module was constructed to create a comprehensive attention‑convolution search space，thereby increasing the network’s focus on discriminative features. Subsequently,a densely connected reduction cell(DCR cell)with more shallow feature input nodes was developed to retain additional shallow feature information，reducing the loss of discriminative feature information and promoting multi‑scale feature fusion.Lastly,the positions of DCR cells were adjusted when stacking the best cells to create network models of varying parameter sizes,enabling deployment on a broader range of terminal devices.The results showed that this method took approximately 4.5 hours to find the optimal neural network model,achieving classification accuracies of 96.14% on the Oxford 102 dataset and 94.12% on the Flower 17 dataset.Compared with methods like AGNAS,it improved accuracy by 1.40 percentage points on Oxford 102 and 3.09 percentage points on Flower 17.

Key words: Neural network architecture search, Convolutional neural network, Attentional mechanism, Fine?grained flower classification

摘要： 为了提升深度卷积神经网络设计的自动化程度，并进一步提高细粒度花卉图像的分类准确率，提出了一种改进的基于DARTS的神经网络搜索方法，用于自动构建细粒度花卉图像分类模型。首先，通过构建注意力-卷积模块，形成全注意力-卷积搜索空间，增强网络对可判别特征的关注度。其次，通过构建具有更多浅层特征输入节点的密集连接缩减单元（DCR cell），保留更多的浅层特征信息，减少可判别特征信息的损失并促进多尺度特征融合。最后，在堆叠最佳cell时调整DCR cell的位置，构建参数量大小不一的网络模型，以便在更多的终端设备上部署。结果表明，该方法耗时4.5 h搜索到了最佳神经网络模型，在Oxford 102和Flower 17上的分类准确率分别为96.14%和94.12%。与AGNAS等方法相比，在Oxford 102上提高了1.40百分点，在Flower 17上提高了3.09百分点。

关键词: 神经网络架构搜索, 卷积神经网络, 注意力机制, 细粒度花卉分类

CLC Number:

S126

ZHENG Xingkai, YANG Tiejun, HUANG Lin. Fine‑Grained Flower Image Classification Based on Neural Network Architecture Search[J]. Journal of Henan Agricultural Sciences, 2024, 53(5): 164-171.

郑兴凯, 杨铁军, 黄琳. 基于神经网络架构搜索的细粒度花卉图像分类方法研究[J]. 河南农业科学, 2024, 53(5): 164-171.

References

［1］ KRIZHEVSKY A，SUTSKEVER I，HINTON G E. ImageNet classification with deep convolutional neuralnetworks［J］. Communications of the ACM，2017，60（6）： 84‑90.

［2］ SIMONYAN K，ZISSERMAN A. Very deep convolutionalnetworks for large‑scale image recognition［EB/OL］.（2015‑04‑10）［ 2023‑05‑10］. https：//arxiv. org/abs/ 1409. 1556.

［3］ SZEGEDY C，LIU W，JIA Y Q，et al. Going deeper with convolutions［C］//2015 IEEE Conference on ComputerVision and Pattern Recognition（CVPR）. Boston，MA，USA：IEEE，2015：1‑9.

［4］ HE K M，ZHANG X Y，REN S Q，et al. Deep residual learning for image recognition［C］//2016 IEEE Conference on Computer Vision and Pattern Recognition（CVPR）. Las Vegas，NV，USA：IEEE，2016：770‑778.

［5］李伟豪，詹炜，周婉，等.轻量型 Yolov7-TSA网络在茶叶病害检测识别中的研究与应用［J］.河南农业科学， 2023，52（5）：162‑169.

LI W H，ZHAN W，ZHOU W，et al. Research and application of lightweight Yolov7‑TSA network in teadisease detection and identification［J］.Journal of Henan Agricultural Sciences，2023，52（5）62‑169.

［6］王川，赵恒滨，李国强，等.基于改进 YOLOX模型的芝麻蒴果检测方法研究［ J］.河南农业科学， 2022，51（11）：155‑162.

WANG C，ZHAO H B，LI G Q，et al. Detection method of sesame capsules based on improved YOLOX model［J］.Journal of Henan Agricultural Sciences，2022，51（11）：155‑162.

［7］任意平，夏国强，李俊丽 .基于优化 AlexNet的花卉识别［J］.电子测量技术， 2020，43（19）：94‑98.

REN Y P，XIA G Q，LI J L.Flower recognition based on optimized AlexNet［J］.Electronic Measurement Technology，2020，43（19）4‑98.

［8］熊举举，徐杨，范润泽，等.基于轻量化视觉 Transformer的花卉识别［J］.图学学报， 2023，44（2）：271‑279. XIONG J J，XU Y，FAN R Z，et al. Flowers recognitionbased on lightweight visual transformer［J］. Journal of Graphics，2023，44（2）：271‑279.

［9］ CAO S，SONG B. Visual attentional‑driven deep learning method for flower recognition［J］. Mathematical Biosciences and Engineering，2021，18（3）：1981‑1991.

［10］ CLBUK M，BUDAK U，GUO Y H，et al. Efficient deepfeatures selections and classification for flower speciesrecognition［J］. Measurement，2019，137：7‑13.

［11］ PEREIRA‑FERRERO V H，VALEM L P，EDRONETTE D C G. Feature augmentation based on manifold rankingand LSTM for image classification［J］. Expert Systems with Applications，2023，213：118995.

［12］ ELSKEN T，METZEN J H，HUTTER F. Neural architecture search：A survey［J］. Journal of Machine Learning Research，2019，20（55）：1‑21.

［13］ ZOPH B，LE Q V. Neural architecture search withreinforcement learning［EB/OL］.（2017‑02‑15）［2023‑05‑10］. https：//arxiv. org/abs/1611. 01578.

［14］ HSU C H，CHANG S H，LIANG J H，et al. MONAS： Multi‑objective Neural Architecture Search using reinforcement learning［EB/OL］.（2018‑12‑03）［2023‑05‑10］. https：//arxiv. org/abs/1806. 10332.

［15］ LIU H X，SIMONYAN K，VINYALS O，et al. Hierarchical representations for efficient architecturesearch［EB/OL］.（2018‑02‑22）［2023‑05‑10］. http：// arxiv. org/abs/1711. 00436.

［16］ ZHONG Z，YANG Z C，DENG B Y，et al. Blockqnn： Efficient Block‑wise neural network architecture generation［J］. IEEE Transactions on Pattern Analysisand Machine Intelligence，2021，43（7）：2314‑2328.

［17］ PHAM H，GUAN M Y，ZOPH B，et al.Efficient neural architecture search via parameters sharing［C］//2018 International Conference on Machine Learning. Stockholm，Sweden：PMLR，2018：4095‑4104.

［18］ BROCK A，LIM T，RITCHIE J M，et al. SMASH： One‑shot Model Architecture Search through HyperNetworks［EB/OL］.（2017‑08‑17）［2023‑05‑10］. https：//arxiv. org/abs/1708. 05344.

［19］ LIU H X，SIMONYAN K，YANG Y M. DARTS：Differentiable architecture search［EB/OL］.（2019‑04‑23）［2023‑05‑10］. http：//arxiv. org/abs/1806. 09055.

［20］ FANG J M，SUN Y Z，ZHANG Q，et al. Densely connected search space for more flexible neural architecture search［C］//2020 IEEE/CVF Conferenceon Computer Vision and Pattern Recognition（CVPR）. Seattle，WA，USA：IEEE，2020：10625‑10634.

［21］ HU J，SHEN L，SUN G. Squeeze‑and‑excitation networks［C］//2018 IEEE/CVF Conference on Computer Visionand Pattern Recognition. Salt Lake City，UT，USA：IEEE，2018：7132‑7141.

［22］ WOO S，PARK J，LEE J Y，et al. CBAM：convolutional block attention module［C］//2018 European Conferenceon Computer Vision. Tel Aviv，Israel：Springer，2018： 3‑19.

［23］ WANG Q L，WU B G，ZHU P F，et al. ECA‑net： Efficient channel attention for deep convolutional neural networks［C］//2020 IEEE/CVF Conference onComputer Vision and Pattern Recognition（CVPR）. Seattle，WA，USA：IEEE，2020：11531‑11539.

［24］ SZEGEDY C，VANHOUCKE V，IOFFE S，et al. Rethinking the inception architecture for computer vision［C］//2016 IEEE Conference on Computer Visionand Pattern Recognition（CVPR）. Las Vegas，NV，USA： IEEE，2016：2818‑2826.

［25］ HOWARD A G，ZHU M L，CHEN B，et al. MobileNets： Efficient convolutional neural networks for mobile vision applications［EB/OL］.（2017‑04‑17）［ 2023‑05‑10］. http：//arxiv. org/abs/1704. 04861.

［26］ XU Y H，XIE L X，ZHANG X P，et al. PC‑DARTS： Patial channel connections for memory efficient differentiable architecture search［EB/OL］.（2020‑04‑07）［2023‑05‑10］. http：//arxiv. org/abs/1907. 05737.

［27］ CHU X X，ZHANG B. Noisy differentiable architectu‑re search［EB/OL］.（2021‑10‑17）［2023‑05‑10］. https：// arxiv. org/abs/2005. 03566.

［28］ SUN Z，HU Y，LU S，et al. AGNAS：Attention‑guided micro and macro‑architecture search［C］//2022 International Conference on Machine Learning. Baltimore，USA：PMLR，2022：20777‑20789.

［29］尹红，符祥，曾接贤，等.选择性卷积特征融合的花卉图像分类［ J］.中国图象图形学报， 2019，24（5）： 762‑772. YIN H，FU X，ZENG J X，et al. Flower image classification with selective convolutional descriptor aggregation［J］. Journal of Image and Graphics，2019，24（5）：762‑772.

［30］杨旺功，淮永建 .多层特征融合及兴趣区域的花卉图像分类［ J］.哈尔滨工程大学学报， 2021，42（4）： 588‑594. YANG W G，HUAI Y J. Flower fine‑grained imageclassification based on multilayered feature fusion andregion of interest［J］. Journal of Harbin Engineering University，2021，42（4）：588‑594.

［31］ XIA X L，XU C，NAN B. Inception‑v3 for flower classi‑fication［C］//2017 2nd International Conference on Image，Vision and Computing（ICIVC）. Chengdu，China：IEEE，2017：783‑787.

[1]	WANG Huanxin, SHEN Zhihao, LIU Quan, LIU Jinjiang. Identification of Crop Leaf Diseases Based on Improved MobileNetV2 Model [J]. Journal of Henan Agricultural Sciences, 2023, 52(4): 143-151.
[2]	. Identification of Farmed Aquatic Animals Types in Freshwater Aquaculture Ponds Based on Sentinel‑2 Imagery [J]. Journal of Henan Agricultural Sciences, 2022, 51(4): 160-170.
[3]	WANG Chuan, ZHAO Hengbin, LI Guoqiang, ZHANG Jiantao, GAO Tongmei, ZHAO Qiaoli, ZHENG Guoqing. Detection Method of Sesame Capsules Based on Improved YOLOX Model [J]. Journal of Henan Agricultural Sciences, 2022, 51(11): 155-162.
[4]	FENG Xiao, LI Dandan, WANG Wenjun, ZHENG Guoqing, LIU Haijiao, SUN Yongsheng, LIANG Shan, YANG Ying, ZANG Hecang, ZHANG Hui, . Image Recognition of Wheat Leaf Diseases Based on Lightweight Convolutional Neural Network and Transfer Learning [J]. Journal of Henan Agricultural Sciences, 2021, 50(4): 174-180.
[5]	XUE Yong, WANG Liyang, ZHANG Yu, SHEN Qun. Recognition Method of Bee Collecting Pollen Behavior Based on Convolutional Neural Network [J]. Journal of Henan Agricultural Sciences, 2020, 49(8): 162-172.