Deep learning: An Image Classification Bootcamp

Build a neural network and image classifier with TensorFlow in Google Colab, using Python and basic machine learning concepts, training on cats, dogs, and the fashion amnesty dataset.

Explore how machine learning turns labeled data into rules and patterns, then build a basic neural network that yields a computer vision model capable of identifying different objects.

Explore the fundamentals of a neural network with input, hidden, and output layers, weights and biases, activation functions like ReLU and sigmoid, and regularization to prevent overfitting in image classification.

Use Google Colab as a free cloud service to boost your Python skills and develop deep learning apps with pre-installed libraries and free DPA for faster training.

Learn how a neural network uses TensorFlow in Python to train with data, loss, and an optimizer across epochs, improving predictions from X to Y.

Explore the cat and dog image dataset with training and test folders, and normalize RGB pixel values from 0-255 to 0-1 for a neural network that distinguishes cats from dogs.

Use the image data generator in TensorFlow to load labeled images from folders, resize at runtime, and feed batched training and validation data for a binary classifier (cats and dogs).

Build an image classifier to distinguish cats and dogs using a dataset loaded from Google Drive, with training and validation images, exploring hidden layers and an optimizer.

Learn how convolutions with filters across pixel neighborhoods emphasize image features, and combine them with max-pooling to compress the image for robust classification.

Apply convolutional layers with 64 3x3 filters and max pooling to distinguish cats from dogs, using padding to preserve image size and improve accuracy on unseen data.

Build a multi-class image classifier using the amnesty version dataset of fashion items. Use softmax activation to output class probabilities and distinguish items like shoes, bags, and caps.

Develop a multiclass image classifier using a 10-class, 28x28 grayscale dataset named fashion MNIST-style, achieving about 0.9 accuracy and discussing loss functions and real-world image considerations.

Create a multi-class image classifier using a handwritten dataset, such as hot dog vs not hot dog, in Google Colab, train a model, and share code with the class.

Learn the basic components of a neural network, build a basic neural network with TensorFlow, apply cnn with max pooling for multiclass classification, using the M9 dataset.