RetinaFace-Emotion-Detection

Facial Expression Recognition with RetinaFace as face detector and Region of Interest (RoI) based ResNet as Emotion Detector

Docs

• Report - fa
• Slides - en

Wrapper on our other repositories

• Facial-Landmark-Detection
• RoI-ResNet
• FaceNet-FER2013
• ResNet50-CIFAR10-FER2013
• AffectNet-Trainer

Classes

1. Angry
2. Disgust
3. Fear
4. Happy
5. Sad
6. Surprise
7. Neutral

Model Checkpoints

You need to place checkpoints in checkpoints folder.

• RetinaFace model weights: [Download]
• Dlib landmark detector model weights [Download]
• RoI ResNet56 trained on FER2013 model weights [Download]

How to run

Change directory to code folder

cd code

Install dependencies

conda env create -f environment.yml

conda activate retinaface-roi

Image and results will be saved in results folder

python run.py --img_path="../photos/test1.jpg"

Webcam

python run.py --webcam=True

Results

Input

Detected faces using RetinaFace

RoI Mask

Output (from left to right)

1. Happy
2. Happy
3. Neutral
4. Surprise
5. Surprise

Webcam Demo

python run.py --webcam=True

References

https://github.com/peteryuX/retinaface-tf2