Update README.md

Dataset:
- Location: 'dataset_even' directory with train, validation, and test splits
- Description: Binary classification dataset with images of size 150 x 150 pixels
Training Process:
- Framework: TensorFlow / Keras
- Architecture: Transfer learning with VGG16 as base model
- Optimizer: Adam with learning rate of 0.0005
- Loss Function: Binary crossentropy
- Batch size: 32
- Number of epochs: Up to 100 with early stopping
Data Preprocessing:
- Rescaling: All images rescaled to 1/255
- Data Augmentation (training set only):
- Rotation range: 20 degrees
- Width shift range: 0.2
- Height shift range: 0.2
- Horizontal flip: Enabled
- Zoom range: 0.2
- Shear range: 0.2
Model Architecture:
- Base: VGG16 (pre-trained on ImageNet, without top layers)
- Additional layers:
Flatten
Dense (256 units, ReLU activation, L2 regularization)
Batch Normalization
Dropout (0.5)
Output Dense (1 unit, sigmoid activation)
- Training Details:
Transfer Learning: Last 4 layers of VGG16 unfrozen for fine-tuning
Early Stopping: Monitoring validation loss with patience of 10 epochs
Class weights: Balanced (1:1)
- Model Performance:
Test Accuracy: [Value not provided in the code, needs to be filled in]
Precision: [Value not provided in the code, needs to be filled in]
F1 Score: [Value not provided in the code, needs to be filled in]
- Additional Metrics:
A full classification report is generated, which would include precision, recall, and F1-score for each class

README.md

@@ -5,6 +5,9 @@ language:
 metrics:
 - accuracy
 pipeline_tag: token-classification
+base_model:
+- timm/vgg16.tv_in1k
+training process: null
 ---
 ---
 license: openrail