This project involves the creation of a food vision model using TensorFlow and EfficientNetB0 as the base model. The model is trained on the CIFAR-100 dataset, which contains 100 classes of images.
The goal of this project is to build a deep learning model capable of classifying food images into different categories. We used the EfficientNetB0 architecture for feature extraction and performed fine-tuning to improve the model's accuracy.
We utilized the CIFAR-100 dataset, which consists of 60,000 32x32 color images in 100 classes, with 600 images per class. There are 50,000 training images and 10,000 test images.
The model is based on the EfficientNetB0 architecture, a state-of-the-art convolutional neural network model that balances accuracy and computational efficiency. Below is the summary of the model:
| Layer (type) | Output Shape | Param # |
|---|---|---|
| input_layer (InputLayer) | (None, 224, 224, 3) | 0 |
| efficientnetb0 (Functional) | (None, 7, 7, 1280) | 4,049,571 |
| global_avtaging_pooling_2d | (None, 1280) | 0 |
| output_layer (Dense) | (None, 101) | 129,381 |
| softmax_float32 (Activation) | (None, 101) | 0 |
- Total params: 12,452,816 (47.50 MB)
- Trainable params: 4,136,929 (15.78 MB)
- Non-trainable params: 42,023 (164.16 KB)
- Optimizer params: 8,273,864 (31.56 MB)
The model was trained for a total of 8 epochs, including 5 epochs of fine-tuning the top layers. Below are the final training metrics:
- Training Accuracy: 99.44%
- Training Loss: 0.0293
- Validation Accuracy: 83.33%
- Validation Loss: 0.8154
- Initial Training: The model was initially trained for 3 epochs with the base layers frozen.
- Fine-Tuning: The top layers of the EfficientNetB0 model were unfrozen, and the entire model was fine-tuned for 5 additional epochs.
- Optimizer: Adam
- Learning Rate: Adaptive learning rate with a scheduler.
- Loss Function: Sparse Categorical Crossentropy
The model achieved a high training accuracy of 99.44% and a validation accuracy of 83.33% after fine-tuning. Although there is some overfitting indicated by the difference between training and validation accuracy, the model shows promising results for food image classification tasks.
To run this project, you can use Google Colab. Here's how to get started:
- Open Google Colab: Google Colab
- Upload the notebook (
.ipynbfile) or clone the repository. - Make sure to select a GPU runtime by navigating to
Runtime>Change runtime typeand selectingGPUunder the hardware accelerator. - Run the notebook cells to train and evaluate the model.
If you'd like to contribute to this project, feel free to fork the repository and submit a pull request. All contributions are welcome!
This project is licensed under the GNU GENERAL PUBLIC License V3. See the LICENSE file for details.