https://github.com/talkuhulk/emotion_classification

emotion classification using fer2013 datasets with a Tensorflow CNN model.
https://github.com/talkuhulk/emotion_classification

classification cnn fer2013 resnet vgg

Last synced: about 1 month ago
JSON representation

emotion classification using fer2013 datasets with a Tensorflow CNN model.

• Host: GitHub
• URL: https://github.com/talkuhulk/emotion_classification
• Owner: TalkUHulk
• Created: 2019-08-05T07:11:40.000Z (over 5 years ago)
• Default Branch: master
• Last Pushed: 2022-11-21T21:31:57.000Z (about 2 years ago)
• Last Synced: 2023-03-05T13:19:13.009Z (almost 2 years ago)
• Topics: classification, cnn, fer2013, resnet, vgg
• Language: Python
• Size: 32.4 MB
• Stars: 18
• Watchers: 1
• Forks: 11
• Open Issues: 4
• Metadata Files:
  • Readme: README.md

Awesome Lists containing this project

README

        
# emotion_classification

This is my personal exercises that use fer2013 datasets with Tensorflow to classify face emotion.

I used vgg and resnet respectively,but all perform poor.In the test datasets,two model achieve accuracy of 54% and 48%.

Except the reasons I do incorrectly, there are several problems with this datasets:

1. Pixel is too low

2. Have a front face, side face, cartoon face and so on. Some kinds of faces that I can't tell what.

3. Data is not a face.

The pre-trained model be put into `model/` folder.

# Below picture shows the training-process on tensorboard:

Vgg:



   


   Loss 





   


   Accuracy



Resnet:



   


   Loss 





   


   Accuracy



# camera/video demo: