摘要
本文以中国菜作为研究对象,提出了基于双线性模型的菜品识别方法.由于中国菜里很多菜品的相似性,导致分类的难度很大.本文借鉴了细粒度图像识别方法中的双线性模型,然后使用一种基于映射的方法得到一个更加低维的双线性特征表示.并在训练阶段采用一种大裕量softmax损失函数.该损失函数通过增加一个正整数变量,在损失函数里产生一个裕量,使同种类别的学习难度增加,从而使学到的特征更加有区分性.将网络在一个208类的中国菜数据集上的测试表明,与以往的方法相比,该方法提高了准确率,减少了过拟合,取得了更好的分类结果.
This article is based on bilinear model,focusing on Chinese food image recognition. The similarity between different Chinese food adds much difficulty for classification. Bilinear cnn model used for fine-grained classification is proved to be helpful in our experiment.Then a new projection algorithm is adopted to compress the high dimension of the bilinear vector. Moreover,this article implements a new loss function during training,which uses a large margin to make the samples in the same class converge harder. Validation on a 208 chinesefood dataset indicates that these two method integrated achieve a better result,not only improving the accuracy,but reducing overfitting.
引文
[1] Lin T Y,RoyChowdhury A,Maji S. Bilinear cnn models for finegrained visual recognition[C]. Proceedings of the IEEE International Conference on Computer Vision,2015:1449-1457.
[2]Gao Y,Beijbom O,Zhang N,et al. Compact bilinear pooling[C].Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition,2016:317-326.
[3]Liu W,Wen Y,Yu Z,et al. Large-margin softmax loss for convolutional neural netw orks[C]. International Conference on M achine Learning,2016:507-516.
[4]Yang S,Chen M,Pomerleau D,et al. Food recognition using statistics of pairwise local features[C]. Computer Vision and Pattern Recognition(CVPR),2010 IEEE Conference on,IEEE,2010:2249-2256.
[5]Anthimopoulos M M,Gianola L,Scarnato L,et al. A food recognition system for diabetic patients based on an optimized bag-of-features model[J]. IEEE Journal of Biomedical and Health Informatics,2014,18(4):1261-1271.
[6]Hassannejad H,Matrella G,Ciampolini P,et al. Food image recognition using very deep convolutional netw orks[C]. Proceedings of the 2nd International Workshop on M ultimedia Assisted Dietary M anagement,ACM,2016:41-49.
[7]Kong F,Tan J. Dietcam:regular shape food recognition with a camera phone[C]. Body Sensor Netw orks(BSN),2011 International Conference on,IEEE,2011:127-132.
[8]Wen Y,Zhang K,Li Z,et al. A discriminative feature learning approach for deep face recognition[C]. European Conference on Computer Vision,Springer,Cham,2016:499-515.
[9]Chen X,Zhu Y,Zhou H,et al. ChineseFoodNet:a large-scale image dataset for Chinese food recognition[J]. arX iv Preprint arX iv:1705. 02743,2017.
[10]Yang Gao-pu,Huang Lv-chen. The outlet of Chinese fast food under the impact of w estern fast food[J]. Shopping M all M odernization,2017,(11):256-257.
[11]Pham N,Pagh R. Fast and scalable polynomial kernels via explicit feature maps[C]. Proceedings of the 19th ACM SIGKDD International Conference on Know ledge Discovery and Data M ining,ACM,2013:239-247.
[12]Kong F,Tan J. DietCam:automatic dietary assessment with mobile camera phones[J]. Pervasive and M obile Computing,2012,8(1):147-163.
[13]Martinel N,Foresti G L,Micheloni C. Wide-slice residual networks for food recognition[C]. Applications of Computer Vision(WACV),2018 IEEE Winter Conference on,IEEE,2018:567-576.
[14]Ciocca G,Napoletano P,Schettini R. Food recognition:a new dataset,experiments,and results[J]. IEEE Journal of Biome,2017,21(3):588-589.
[15]Xu R,Herranz L,Jiang S,et al. Geolocalized modeling for dish recognition[J]. IEEE Transactions on Multimedia,2015,17(8):1187-1199.
[16]Herranz L,Jiang S,Xu R. Modeling restaurant context for food recognition[J]. IEEE Transactions on Multimedia,2017,19(2):430-440.
[17]Zhang N,Donahue J,Girshick R,et al. Part-based R-CNNs for finegrained category detection[C]. European Conference on Computer vision,Springer,Cham,2014:834-849.
[10]杨镐璞,黄履晨.西式快餐冲击下中式快餐的出路[J].商场现代化,2017,(11):256-257.