Comparing Behaviors of Various CNN Models When Applied to the Car Model Classification Problem

Yixuan Wang ； Hanfang Xue ； Jie Deng ； Yuran Bi ； Zihao Zhang

《World Scientific Research Journal》 7卷1期 (2021/01) Pp. 250-260

https://doi.org/10.6911/WSRJ.202101_7(1).0034

摘要

In this paper, an experiment comparing various deep learning models based on their performances on the car classification problem was conducted. Deep learning models ResNet-18, ResNet-34, ResNet-50, ResNet-101, ResNet-152, MobileNet V2 and DenseNet-121 was included in our experiment. The Stanford Cars dataset from Kaggle was used, and it was further expanded by our Selenium Python crawler. Indeed, classifying car models is a fine-grained classification problem, which was coped with by scaling the loss while training. Accuracies and confusion matrices were calculated for each model for evaluation. For further interpretation, Grad-CAMs was used to analyze these models, which proved the reliability of the models. The results indicate that most of the resultant models work well, can be interpreted and explained with Grad-CAMs, and could be deployed in real life for various purposes, such as car information apps.

關鍵字

Car Model Classification ； CNN ； Fine-grained Classification ； Grad-CAMs

參考文獻

K. He, X. Zhang, S. Ren, and J. Sun, ‘Deep Residual Learning for Image Recognition', in 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Las Vegas, NV, USA, Jun. 2016, pp. 770–778, doi: 10.1109/CVPR.2016.90.

G. Huang, Z. Liu, L. Van Der Maaten, and K. Q. Weinberger, ‘Densely Connected Convolutional Networks', in 2017 IEEE Conference on Computer Vision and Pattern Recognition (CVPR), Honolulu, HI, Jul. 2017, pp. 2261–2269, doi: 10.1109/CVPR.2017.243.

Ramprasaath R. Selvaraju, Michael Cogswell, “Grad-CAM: Visual Explanations from Deep Networks via Gradient-based Localization,” International Journal of Computer Vision, Sep.2019

A. G. Howard et al., ‘MobileNets: Efficient Convolutional Neural Networks for Mobile Vision Applications’, p. 9.

Jesús Utrera, Stanford Car Dataset by classes folder, Jul.2018 https://www.kaggle.com/jutrera/stanford-car-dataset-by-classes-folder