FaceShape / README.md
fatyidha's picture
Update README.md
ca915d8 verified
---
library_name: keras
tags:
- tensorflow
- keras
- face-shape-classification
- cnn
pipeline_tag: image-classification
license: apache-2.0
---
# FaceShape Model
## Model Description
This model classifies facial shapes into categories such as oval, square, round, etc. It is designed for applications in virtual try-ons and eyeglass frame recommendations.
- **Framework:** TensorFlow (Keras)
- **Model Format:** `.h5`
- **Purpose:** Face shape classification.
## How to Use
To use this model, you can load it with TensorFlow and Keras. Below is an example:
```python
from tensorflow.keras.models import load_model
# Load the model
model = load_model("path_to_your_model.h5")
# Example input
input_data = ... # Replace with your preprocessed input
output = model.predict(input_data)
print(output)
## Training Details
The model was trained using a Convolutional Neural Network (CNN) architecture on the [Face Shape Classification Dataset](https://www.kaggle.com/datasets/lucifierx/face-shape-classification).
### Preprocessing Steps
- **Image Size**: All input images were resized to 224x224 pixels.
- **Normalization**: Pixel values were normalized to the range [0, 1].
- **Data Augmentation**: Techniques like rotation, flipping, and zooming were applied to improve generalization.
### Training Configuration
- **Framework**: TensorFlow (Keras)
- **Optimizer**: Adam
- **Loss Function**: Categorical Crossentropy
- **Batch Size**: 32
- **Epochs**: 50
- **Validation Accuracy**: Achieved 85% on the validation set.
### Hardware
The model was trained on an NVIDIA GPU for faster computation.
## Limitations
- The model may not perform well with low-resolution or occluded images.
- The dataset may not represent all possible face shapes, which could limit generalization.
## Example Predictions
Here are some example predictions:
| Input Image | Predicted Class |
|--------------------|-----------------|
| ![example1](https://via.placeholder.com/100) | Oval |
| ![example2](https://via.placeholder.com/100) | Square |