| # Fashion MNIST Classifier Zoo 👕🎽👖 | |
| ## Model Overview | |
| Welcome to the Fashion MNIST Classifier Zoo! This model card showcases a collection of image classification models trained on the [Fashion MNIST dataset](https://github.com/zalandoresearch/fashion-mnist). Each model offers a unique approach to identifying articles of clothing from 28x28 grayscale images. Explore the models below to find the perfect fit for your fashion needs! | |
| ## Models | |
| ### CNN_Fashion_MNIST | |
| - **Architecture:** A custom Convolutional Neural Network (CNN) designed for efficient feature extraction and classification. | |
| - **Size:** 1.5 MB | |
| - **Use Case:** Ideal for resource-constrained environments or applications requiring fast inference. | |
| ### VGG16_Fashion_MNIST | |
| - **Architecture:** Implementation of the classic VGG16 architecture, leveraging its deep layers for robust feature learning. | |
| - **Size:** 184 MB | |
| - **Use Case:** Suitable for applications where high accuracy is paramount, even at the cost of increased computational complexity. | |
| ### Xception_Fashion_MNIST | |
| - **Architecture:** Employs the Xception architecture, known for its efficient use of parameters and strong performance. | |
| - **Size:** 279 MB | |
| - **Use Case:** A good balance between accuracy and computational efficiency, making it suitable for a wide range of applications. | |
| ## Interactive Demo | |
| Unfortunately, this model card is static, but imagine the possibilities! If this were interactive, you could: | |
| 1. **Upload your own fashion images** and see how each model classifies them. | |
| 2. **Compare the models' performance** on a held-out test set with interactive visualizations. | |
| 3. **Adjust confidence thresholds** to explore the trade-off between precision and recall. | |
| ## Intended Use | |
| These models are intended for: | |
| - Educational purposes: Learning about image classification and deep learning architectures. | |
| - Benchmarking: Comparing the performance of different models on the Fashion MNIST dataset. | |
| - Inspiration: Providing a starting point for building more sophisticated fashion recognition systems. | |
| ## How to Use | |
| 1. **Load the Model:** Use TensorFlow/Keras to load the `.keras` model file of your choice. | |
| ``` | |
| from tensorflow import keras | |
| model = keras.models.load_model('VGG16_Fashion_MNIST.keras') | |
| ``` | |
| 2. **Prepare Your Data:** Ensure your input data consists of 99x99x3 RGB images, preprocessed to match the model's expected input. | |
| 3. **Make Predictions:** Use the loaded model to predict the class of each image. | |
| ``` | |
| predictions = model.predict(your_test_data) | |
| ``` | |
| ## Training | |
| The `fashion_mnist.ipynb` notebook provides a complete guide to training these models from scratch. Follow the instructions in the notebook to: | |
| 1. **Load the Fashion MNIST dataset.** | |
| 2. **Preprocess the data.** | |
| 3. **Build and train the models.** | |
| 4. **Evaluate their performance.** | |
| ## Files | |
| - `.gitattributes`: Specifies attributes for files in the repository. | |
| - `CNN_Fashion_MNIST.keras`: Pre-trained CNN model. | |
| - `VGG16_Fashion_MNIST.keras`: Pre-trained VGG16 model. | |
| - `Xception_Fashion_MNIST.keras`: Pre-trained Xception model. | |
| - `fashion_mnist.ipynb`: Jupyter Notebook for training and evaluation. | |
| - `README.md`: This model card. | |
| ## Limitations and Future Directions | |
| - **Dataset Bias:** The Fashion MNIST dataset is a simplified representation of real-world fashion images. Models trained on this dataset may not generalize well to more complex scenarios. | |
| - **Limited Architectures:** This collection includes only a few popular architectures. Future work could explore more recent and advanced models. | |
| - **No Interactive Demo:** As mentioned above, an interactive demo would greatly enhance the user experience. | |
| ## Author | |
| [Harsh Maniya](https://huggingface.co/harshhmaniya) | |
| [GitHUb](https://github.com/harshhmaniya) | |