A small fully connected neural network that classifies the 10 clothing categories of Fashion-MNIST, a set of 28×28 grayscale images.
The notebook loads the data with TensorFlow Datasets, scales the pixels to the range 0 to 1 and batches them in groups of 32. The model is Flatten → Dense(64, ReLU) → Dense(10, softmax), trained with Adam and sparse categorical cross-entropy.
After 5 epochs the model reaches about 87 % accuracy on the test set. A custom Keras callback stops training once the training accuracy passes 90 %.
computer_vision_.ipynb has the code and its outputs. It runs from top to bottom with tensorflow and tensorflow-datasets installed.