Research on Fine-Grain Model Recognition Based on Branch Feedback Convolution Neural Network

EasyChair Preprint 842

Abstract

Fine-grained vehicle identification has a wide range of applications in many fields, and the requirements for recognition accuracy are high in various application scenarios. In this paper, a branch fusion convolutional neural network algorithm for fine-grained car recognition is designed. The VGG16 convolutional neural network is merged with AlexNet to form a bifurcated fusion convolutional neural network. On this basis, the multi-branch training idea of GoogleNet is cited to make the network model stabilize and converge during training. The network was trained and tested on the CompCars fine-grained car dataset. The correct rate of the test set Top-1 reached 91.29%, and the model was accurate and effective.

Keyphrases: Feedback convolutional neural network, VGG16, multi-branch training

@booklet{EasyChair:842,
  author    = {Shijin Li and Ting Shi and Fushou Tao and Rong Jiang},
  title     = {Research on Fine-Grain Model Recognition Based on Branch Feedback Convolution Neural Network},
  howpublished = {EasyChair Preprint 842},
  year      = {EasyChair, 2019}}