Android Machine Learning with TensorFlow lite in Java/Kotlin

Learn how machine learning uses data and algorithms to find patterns, with data sets, training data, testing data, features and labels, and supervised, unsupervised, and reinforcement learning.

Master supervised machine learning by training with labeled data, then testing predictions. Learn regression to predict continuous values and classification for categorical labels, using features and labels.

Explore unsupervised machine learning with unlabeled data and clustering to reveal patterns, illustrated by groceries; and learn reinforcement learning via trial and error with rewards and punishments.

Explore deep learning with artificial neural networks, learning how layered neurons perform feedforward and backpropagation to train models for regression and classification.

Explore core machine learning concepts, including supervised and unsupervised learning, deep learning with neural networks, and key training processes like feed forwarding, back propagation, learning rate, and loss functions.

Learn to use Google Colab to run Python in your browser with zero configuration, access free GPUs, and train and evaluate machine learning models using NumPy and Matplotlib.

Explore python's basic syntax and data types, including numbers and strings, declare variables without explicit types, and perform simple operations in Google Colab for training ml models on mobile apps.

Master Python lists as versatile containers that hold mixed types in square brackets, enable index-based access and printing, and support methods like append, insert, remove, sort, and reverse.

Explore Python data structures by comparing tuples to lists—tuples use parentheses, are immutable, and support indexing and slicing—then learn dictionaries with key–value pairs and key-based retrieval.

Explore Python conditional statements with if, elif, and else to compare two values, and practice while and for loops to iterate lists and strings, plus handling user input.

Open and close files in Python, use read and write modes (r, w, wb, a), and create or verify the model file during ML training.

Learn numpy essentials in Python, creating arrays, exploring shapes and dtypes, and using Colab notebooks to prep data for training mobile image classification models with pandas and matplotlib.

Explore numpy operations to create multi-dimensional arrays using zeros, ones, random, and full; reshape arrays and flatten them for streamlined data preparation in machine learning.

Explore numpy functions by applying unary and binary operators to arrays, perform dot products and matrix multiplication with np.dot and np.matmul, and sort multidimensional arrays by axis.

Discover the pandas data science library for manipulation in Python, and learn to load data into data frames, create frames from lists and dictionaries, and work with rows and columns.

Load a csv file with pandas read_csv to create a data frame from a comma separated values dataset, and upload it to Google Colab specifying its path and index column.

Learn how to load a csv into a pandas data frame, detect missing values with isnull, fill or drop them with fillna and dropna, and inspect shape for model training.

Explore how to plot data with matplotlib in Python using plt.plot, plt.bar, and plt.scatter to visualize x and y arrays in notebooks.

Load and display images using Matplotlib in Python, treating images as multi-dimensional arrays and exploring their shape and index-based access as with a Python list.

Learn to declare and manipulate variables and constants in TensorFlow, using tf.variable and tf.constant, reassign with assign, and control data types and shapes for mobile machine learning with TensorFlow Lite.

Learn to shape and reshape tensors, determine their rank, access elements by index, and convert to numpy, while using tf.size and dtype checks.

Explore element-wise operations in TensorFlow, including multiply, add, subtract, and divide, and learn matrix multiplication with transposed operands. Understand typecasting and rag tensor, where rows have different lengths.

Learn how to compute tensor means with tf.reduce_mean, specifying axis for columns or rows, and generate tensors with random values using TensorFlow.

Discover how to generate random values in TensorFlow using normal and uniform distributions, control shape and dtype, set seeds for repeatable results, and concatenate tensors.

Learn how TensorFlow checkpoints save and restore variable values during model training, enabling resume after failures and efficient storage of hundreds of thousands of values.

Train a simple linear regression model to uncover the pattern output equals two times input minus one from a two-column dataset, then test it in a Flutter app.

Train a first linear regression model from a simple x to y=2x-1 dataset using TensorFlow and Keras, then convert the trained model to tflite for mobile apps.

Test trained model with predict, confirming outputs follow two times input minus one, then convert it to tflite by saving to h5 and using tf lite converter for mobile.

Review training a first linear regression model with TensorFlow, using a simple one-hidden-layer, one-neuron network, and converting it to a tflite file for mobile applications.

Analyze the tflite model to inspect input and output shapes using natron, revealing a single input value and a single output value with float 32 for mobile app integration.

Create a new android project and design a GUI with input, a button, and a result view to run a regression model that outputs two times the input minus one.

Integrate a TensorFlow Lite model into an Android project by adding an assets folder, placing the tflite file, syncing the TensorFlow dependency, and loading the model with a Java/Kotlin interpreter.

Pass user input from a text field to the tflite interpreter, convert it to float, run inference, and display the output in a TextView.

Learn how to integrate a tflite regression model into an Android app by adding the model to assets, configuring gradle, loading with the tflite interpreter, and running inference.

Train your first real world regression model to predict mpg (fuel efficiency) for vehicles using features like cylinders, displacement, weight, and horsepower from the auto mpg dataset on Kaggle.

Load the fuel efficiency dataset in Google Colab using pandas read_csv into a data frame, train a TensorFlow Keras neural network for regression, and convert the model to tflite.

Identify missing values in the dataset and handle them by replacing with a value or dropping affected rows using dropna, ensuring clean data for model training.

Learn to handle categorical columns using one hot encoding by converting the origin category into separate United States, Europe, and Japan features, removing origin column, and boosting model accuracy.

Split the dataset into 80% training and 20% testing with sample and draw, separate miles per gallon as the label, and inspect stats with describe for normalization and model training.

Normalize dataset features by subtracting the mean and dividing by the standard deviation for each column, applying this to training and test data before model training.

Normalize training and test data, then build and train a Keras sequential neural network in TensorFlow Lite for fuel efficiency prediction.

Test a trained model with a 20% test split, compare predictions to fuel efficiency values, compute loss (mae and mse), and convert the Keras model to tflite for Android apps.

Train a real-world regression model for fuel efficiency using a Kaggle dataset in Google Colab, apply one-hot encoding and normalization, split 80/20, and convert to tflite for Android apps.

Import the starter Android project from GitHub, install it in an emulator, and input vehicle data (cylinders, displacement, horsepower, weight, acceleration, model year, region) to predict and display fuel efficiency.

Examine the starter Android app code, including MainActivity.java and the XML layout, and learn how the GUI uses a spinner with USA, Europe, and Japan and a TensorFlow Lite model.

Load a TensorFlow Lite model in an Android app with a mapped ByteBuffer and interpreter, then prepare and normalize UI inputs (cylinders, displacement, horsepower, weight, acceleration, year, region one-hot) for inference.

Learn to normalize Android app inputs for a TensorFlow Lite model by subtracting feature means and dividing by standard deviations using train stats, applying to nine input values.

Pass nine input values in a 2D float array to a tflite model in Android, run inference with the interpreter, and read a single output value for mpg.

Test the fuel efficiency prediction Android app in the simulator, comparing the predicted 15.02 with the actual 15 using cylinders, displacement, horsepower, weight, acceleration, model year, and origin USA.

Learn how to build a fuel efficiency prediction Android app by loading a TensorFlow Lite model from assets, normalizing inputs, and running inference to display the result.

Define a simple function, convert it to a concrete function, and export it as a TensorFlow Lite model for Android using input signatures and a converter.

Learn to save a concrete function as a saved model, convert it to a TensorFlow Lite model using the offload converter, and integrate the exported file into an Android app.

Build an Android app that lets users draw digits with a fingerpaint view, converts to a bitmap, runs it through a TensorFlow lite model, and shows prediction, probability, and time.

Explore transfer learning in android machine learning with TensorFlow Lite by using a model trained on vegetables as a starting point to classify flowers, boosting performance.

Explore Google Colab to train machine learning models in your web browser with a cloud-based notebook. Create, run, download notebooks or Python files, and save to drive.

train and deploy a flower recognition model on Android using TensorFlow Lite, transfer learning, and the model customization library, including preparing data in Google Colab and converting to on-device model.

Train and evaluate a flower recognition model with transfer learning in TensorFlow Lite for Android, fine-tuning only the last layer and exporting the tflite model and labels.

Build an Android flower recognition model with TensorFlow Lite in Java/Kotlin using transfer learning, loading a labeled dataset, training with validation, evaluating with test data, and exporting a tflite model.

Switch to the B0 model for on-device image classification of flower recognition and compare training, validation, and testing accuracies against the MobileNet baseline.

Evaluate a tflite model in an Android app by loading the flight file, creating an interpreter, preprocessing test images, running inference, and calculating accuracy.