This is the code for a fashion article classifier that detects clothing and accessories, from the Zalando Research Fashion MNIST repository.
(Image by: zalandoresearch)
I built a 3-layer feedforward neural network which classifies different clothing / accessory items such as t-shirts, trousers, dresses, sneakers, and bags using the TensorFlow library. The inputs are black and white images of size 28x28, and the output is an integer (corresponding to the target class).
| Set | Classification Accuracy |
|---|---|
| Training | 93.51% |
| Test | 89.35% |
| Set | Accuracy (Fashion MNIST) | Accuracy (MNIST) |
|---|---|---|
| Training | 93.51% | 99.74% |
| Test | 89.35% | 97.74% |
Here are the final hyperparameters and parameters which yielded the accuracies above:
| Hyperparameter / Parameter | Value / Type |
|---|---|
| Weight initialization | Xavier |
| Hidden layers | 2 |
| Hidden units (layer 1) | 128 |
| Activations (hidden layers) | ReLU |
| Output layer | Softmax |
| Minibatch size | 32 |
| Learning rate | 0.0005 |
| Epochs | 16 |
| Optimizer | Adam |
- numpy
- tensorflow
Install dependencies using pip.
The data set was taken from the Fashion MNIST repository by zalandoresearch. The training set contains 60,000 examples and the test set contains 10,000 examples. Each example is a 28x28 grayscale image, and has an associated label from 10 classes.
| Target class | Definition |
|---|---|
| 0 | T-shirt/top |
| 1 | Trouser |
| 2 | Pullover |
| 3 | Dress |
| 4 | Coat |
| 5 | Sandal |
| 6 | Shirt |
| 7 | Sneaker |
| 8 | Bag |
| 9 | Ankle boot |
Run the notebook on a localhost server using jupyter notebook.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent