This project addresses the problem of identifying a driver's emotional state from dashboard camera images. Emotions such as stress, fatigue, and anger can increase driving risk, so the aim is to build dependable facial-expression classifiers that support in-cabin safety systems by recognizing seven core emotions.
The notebooks implementing the CNN and ResNet models are independent and may be run separately from each other.
Note:
There is some overlapping code and functions between the notebooks. For clarity, we chose to separate the workflow for each model.
- If you are running on Google Colab, you should not need to install any additional Python packages.
- We perform learning rate tuning and experimentation, which can take a while to run (e.g., approximately 40 minutes using a Colab T4 GPU runtime).
- Recommendation: Consider skipping this code block or reducing the number of epochs. There is a note in the notebook regarding this.
If you have any questions, please feel free to reach out!