Download PDFOpen PDF in browser

Freshness Identification of Iberico Pork Based on Improved Residual Network and Transfer Learning

EasyChair Preprint 708

8 pagesDate: January 2, 2019

Abstract

In order to improve the accuracy of pork freshness identification, a method for pork freshness identification based on improved residual network and transfer learning was proposed. First of all, the pork freshness was classified into fresh, secondary fresh grade I, secondary fresh grade II, secondary fresh grade III, deteriorated grade I, deteriorated grade II and deteriorated grade III, a total of 7 grades, according to the aerobic plate count, coliform bacteria and pH value of pork combined with national pork food standards. The Resenet-50 model was trained with the AAUSet data set to have the ability to extract image features. Then, the Resenet-50 model was improved using model transferring and model fine-tuning in the following ways: first, replace the full connection and classification layers of the Resenet-50 model with a 3-layer adaptive network; next initialize the improved Resenet-50 model weights using the network parameters trained on the AAUSet; then use LReLU as the activation function of the adaptive network; finally, transfer the knowledge gained by the improved Resnet50 model on the image data set of the pork sample to the task of Iberico pork freshness identification. The images of Iberico pork were preprocessed with rotation and scale-zooming, and then images of the 7 freshness grades were selected, a total of 23427 images forming the sample set. Then, 80% of the samples were randomly selected from the sample set to be used as the training set, and the remaining 20% for the test set. The test results showed that transfer learning could significantly improve the convergence speed and classification performance of the model, and data augmentation could increase the diversity of data, avoiding over-fitting phenomena. The accuracy of classification in transfer learning and data augmentation could reach as high as 94.5%. Moreover, the test process was high real-time, making the test method a more efficient method for classifying pork freshness.

Keyphrases: Image identification, Pork freshness, Transfer Learning, residual network

BibTeX entry
BibTeX does not have the right entry for preprints. This is a hack for producing the correct reference:
@booklet{EasyChair:708,
  author    = {Jiao Jun and Wang Wenzhou and Gu Lichuan},
  title     = {Freshness Identification of Iberico Pork Based on Improved Residual Network and Transfer Learning},
  howpublished = {EasyChair Preprint 708},
  year      = {EasyChair, 2019}}
Download PDFOpen PDF in browser