kimjeonghun91/facial_emotion_recognition

It's a project of facial emotion recognition.

facial_emotion_recognition

Recognize facial emotions in 7 categories: angry, disgust, fear, happy, sad, surprise, neutral.

Dataset

The dataset is provided by a competition, which is quite similar to FER2013 dataset.

Model

The API for face detection is Google's mediapipe API. The model for emotion recognition is a 15-layer (8 convs + 4 pooling + 3 fcs) VGG style network.

Pipeline
├── face detection: mediapipe
└── emotion recognition: vggnet

VGG architecture

Data flow

Upsampling

The upsampling technique is SMOTE.

Usage

Train

If you have a dataset, you can train use training.ipynb

Inference

If you want infer directly, use inference.ipynb

The weights I trained is located in saved_models. The default setting is a voting classifer of a model trained by orginal data and a model trained by upsampled data.

Performance

Result

Video

Images