kmg0157 / drowsy_detection_binary

drowsy_detection_binary

This project is to classify images into drowsiness and normal condition.

The ones used for image preprocessing are haar cascade and mediapipe.

WideResnet tried to create a more accurate classification model by expanding the range of models.

Environment & Package

1. NVIDIA Titan XP (CUDA 10.2)
2. Python 3.8.10
3. Tensorflow-gpu 2.10.0
4. Mediapipe 0.9.0.1
5. Keras 2.10.0

Result

Check the code for the detailed structure of the model.

1. Model 1 (CNN)
2. Model 2 (WideResnet+ReLU)
3. Model 3 (WideResnet+Leaky ReLU)

Test

The model with WideResnet did not achieve the desired level of learning results due to a lack of epochs.