Upload 7 files

1. Facial Drowsiness Detection
- Input: The system starts with a live feed or a captured image of the driver.
- Base Model: A baseline model is initially developed to detect facial features and signs of drowsiness. This model might not be highly accurate and can overfit the training data.
- Transfer Learning: To improve performance, pre-trained models like VGG16, ResNet18, and MobileNetV2 are employed. These models are fine-tuned with our specific dataset to leverage their pre-learned features for better accuracy.
- Best Model: Among the models used in transfer learning, the one with the best performance metrics (such as MobileNetV2) is selected as the final model for facial drowsiness detection. Detection: The best model is used to detect if the driver shows signs of facial drowsiness. If Facial Drowsiness Detected: An alert is triggered indicating that the driver is drowsy.

2. Eye Closure Drowsiness Detection
- Input: Parallel to facial analysis, another model focuses specifically on the driver's eyes to detect if they are closed.
- Base Model: A baseline model is initially developed to differentiate between open and closed eyes.
- Hyperparameter Tuning: To enhance the model’s performance, various hyperparameters (such as different optimizers and dropout rates) are tuned. This helps in achieving better accuracy and generalization.
- Hyperparameter Tuned Model: The model with the best hyperparameter settings is selected for eye closure detection.
- Detection: The hyperparameter tuned model is used to check if the driver's eyes are closed. If Eyes Closed: An alert is triggered indicating that the driver is drowsy.

## Key Points
- Facial Drowsiness Detection uses both a base model and transfer learning with pre-trained models to achieve high accuracy.
- Eye Closure Drowsiness Detection focuses on tuning hyperparameters to find the optimal model for detecting closed eyes.
- This pipeline effectively combines different deep learning techniques to create a comprehensive driver drowsiness detection system.

trained_model_weights_BASE_DATASET1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:56da6b041173dce4f7c86b5d68667f201c009ad4ffbd17a6702755615b3a564c
+size 5686498

trained_model_weights_BASE_DROPOUT_DATASET1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:6648381a9ab22b08c9cf2e2bf1a322bea7ec9a588ca5876eb7793b59b2ed655d
+size 5686626

trained_model_weights_BASE_DROPOUT_EARLYSTOPPING_DATASET1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:c6648cb1705eddc84efb13a59c38752e268132a19d3d9287055d2f4a86d53980
+size 5686850

trained_model_weights_FINAL_DATASET2.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:3dd16a8b110c67ec8054d66d26b0d5752b37e773dfd9e345f01bd2fd1a003249
+size 414894322

trained_model_weights_MOBILENETV2_DATASET1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:fb10073e38613fac815c35596a961d2e6a899d69700bd02e3f963b8cad156516
+size 9153226

trained_model_weights_RESNET18_DATASET1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:7be3e94b25c6499c996eeeebc1b0a3988fce3a13368e5e6da589a3873b8f6d13
+size 44789904

trained_model_weights_VGG16_DATASET1.pth

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:33a18f1e6c66c033508e5adcd5d24e7a8b1641ef9c0271517b83da61a2ba4b2c
+size 537070918