https://github.com/tommia/car-make-classifier

Transfer train VGG16 based model to infer car make from image
https://github.com/tommia/car-make-classifier

Last synced: 3 months ago
JSON representation

Transfer train VGG16 based model to infer car make from image

• Host: GitHub
• URL: https://github.com/tommia/car-make-classifier
• Owner: TommiA
• Created: 2018-09-27T12:26:01.000Z (over 6 years ago)
• Default Branch: master
• Last Pushed: 2018-10-02T05:56:51.000Z (over 6 years ago)
• Last Synced: 2025-01-19T13:56:14.876Z (5 months ago)
• Language: Python
• Size: 54.7 KB
• Stars: 0
• Watchers: 2
• Forks: 0
• Open Issues: 0
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# Car-Make-Classifier

Transfer learning on VGG16 based convolutional network model to infer car make from image. Training code loads the basic VGG16 with imagenet weights leaving the top out. The code adds 512-256-81 classifier NN on top of it as my data contained 81 different car make models, i.e. classes. The trials with my training and testing data with this model reached validation accuracy of ~85%. Proper filtering to improve the training and testing data quality could improve that.

![Alt text](plot_E100_IMG400_BS300_INVLR10000.png?raw=true "Title")