Tutorials for TensorFlow

Hands-on notebooks for learning deep learning with TensorFlow, from the building blocks of neural networks to Vision Transformers, Bayesian deep learning and generative models. Many of them come with a detailed blog post.

A similar repository dedicated to Machine Learning and Data Analysis is here.

Libraries

Normalizing Flows: transforming distributions with TensorFlow Probability

Vision Transformer: how self-attention works

---

🆕 Recent

Topic	Description	Code	Blog	Run
ESRGAN	From GAN to SRGAN to ESRGAN: perceptual loss, the Residual-in-Residual Dense Block and the relativistic discriminator for image super-resolution (paper).	ESRGAN_Concepts.ipynb	Medium
Diffusion Math	Every step of the diffusion loss explained: the forward and reverse processes of denoising diffusion probabilistic models (DDPM paper).	Diffusion	Medium

Computer Vision

Topic	Description	Code	Blog	Run
Conv Basics	Start with the building blocks of deep neural networks and slowly build and understand convolutional neural networks. Example below.	Conv_Basic_with_TensorFlow	–
Image Generator	Preprocess images using Keras flow_from_directory, with augmentation on the fly.	im_generator	–
Multi-Class Classification	Moving from binary classification to multi-class classification.	Multi_Class	–
Transfer Learning	Using VGG16 weights to train a model: 2800 images of cats and dogs for training and 600 for validation.	Transfer_Learning	–
Vision Transformer	Implement the Vision Transformer from scratch in TensorFlow 2.0, with a detailed look at how the attention mechanism actually works.	ViT_TensorFlow	TDS

How Conv and Max Pool layers understand features (training images from the Fashion-MNIST data set):

Kaggle & Applied Projects

Topic	Description	Code	Blog	Run
Cassava Leaf Disease	The Kaggle competition on cassava leaf disease classification. This is the same notebook used for the competition submission.	cassava-classification.ipynb	–
Pneumonia Detection	Pneumonia detection from chest X-rays using the TensorFlow Dataset API. Concepts such as image standardization, class imbalance and the ROC curve as a performance metric were influenced by the DeepLearning.AI AI for Medicine course.	detectpneumonia-inceptionresnetv2-class-imbalance.ipynb	–

Bayesian Deep Learning

Topic	Description	Code	Blog	Run
Bayesian Uncertainty	The two types of uncertainty, aleatoric and epistemic, and how a Bayesian neural network can estimate them. Uses real weather data and TensorFlow Probability.	Bayesian_uncertainty.ipynb	TDS

NLP

Topic	Description	Code	Blog	Run
Word Embeddings	Classifying IMDB movie reviews and visualizing how embeddings cluster positive and negative words.	NLP_Embedding1	–
LSTM + GloVe	Using pretrained GloVe embeddings to separate positive tweets from negative ones (Sentiment140, 1.6M tweets).	NLP_LSTM_Glove	–

Time Series

Topic	Description	Code
Week 1	Introduction to time series data.	Time_SeriesW1
Week 2	From simple linear regression to single-layer and deep neural networks.	Time_SeriesW2
Week 3	RNNs and LSTMs for time series.	Time_SeriesW3
Week 4	Combining LSTM and 1D convolution; predicting sunspots and the minimum temperature of Melbourne.	Time_SeriesW4

Predicting the minimum temperature of Melbourne:

Data Pipelines

Topic	Description	Code	Blog	Run
TF Data API	An introduction to the TF Dataset API and how to use it to build efficient, fast data pipelines.	TF_Data_Intro.ipynb	TDS

Note: some notebooks read Kaggle data sets or local files. When running them in Colab, download or upload the matching data first.