Classification of Fashion products using different neural network based models -- Feedforward, CNN and VGG, etc.
Fashion-MNIST is a dataset of Zalando's article images—consisting of a training set of 60,000 examples and a test set of 10,000 examples. Each example is a 28x28 grayscale image, associated with a label from 10 classes. You can download it here -- https://github.com/zalandoresearch/fashion-mnist
| Target Class | Definition |
|---|---|
| 0 | T-shirt |
| 1 | Trouser |
| 2 | Pullover |
| 3 | Dress |
| 4 | Coat |
| 5 | Sandal |
| 6 | Shirt |
| 7 | Sneaker |
| 8 | Bag |
| 9 | Ankle Boot |
- 2 Layer feedforward Nerual Netwrok
- CNN (1 Conv layer)
- CNN (3 Conv layer)
- CNN (4 Conv layer) + Batch Normalization
- VGG
- VGG + Batch Normalization
Since, my data set only had 60,000 test cases, which is quite less for a deep learning model (for eg: VGG). So, I have used data (image) augmentation technique to increase the number of images and also to improve the quality of images.
- Dropout
- Data Augmentation
Out of all experiments, model with VGG + Batch Normalization performed the best with almost 95% accuracy.
| Model | Accuracy |
|---|---|
| 2 Layer Nerual Netwrok | 88% |
| CNN (1 Conv layer) | 91% |
| CNN (3 Conv layer) | 91% |
| CNN (4 Conv layer) + Batch Normalization | 92% |
| VGG | 93% |
| VGG + Batch Normalization | 94% |