Machine Learning on Google Cloud with AutoML and VertexAI

Explore Vertex AI, Google's unified ml platform, and AutoML to train and deploy models with minimal code across images, videos, text, and tabular data.

Explore AutoML for text data to create three machine learning models: classification, entity extraction, and sentiment analysis.

Train a model for text classification by recognizing patterns in text to categorize it into predefined labels. Explore real-world applications like topic categorization, spam detection, language detection, and medical diagnosis.

We begin with AutoML for text classification, training a model and setting up Vertex dataset. We create AutoML resource and endpoint, then evaluate metrics and online batch predictions.

Prepare and label a text dataset to train an auto ML model on Google Cloud with Happy Moments data, csv or json lines, and labels affection, achievement, exercise, and nature.

Create a text classification single-label dataset in Vertex AI via the gcp console, upload a labeled csv to a cloud storage bucket, and review dataset details.

Train a text classification model in Vertex AI AutoML using a prepared dataset with an 80/10/10 split, then review the model in the registry for online and batch predictions.

Deploy your trained text classification model to an endpoint in Google Cloud Console and test online predictions, then review metrics in the model registry and verify labels affection and bonding.

Learn to build and deploy a text classification model using Google Colab and Vertex AI, covering online and batch prediction, data sourcing, training, and endpoint deployment.

Create and run batch predictions in Vertex AI using Google Colab, building input files and json line references, uploading to Google Cloud Storage, and reviewing confidence scores.

Learn entity extraction and named entity recognition in text, identifying persons, organizations, locations, dates, medical terms, and product names for practical business use.

Train a text entity extraction model with AutoML, using a vertex dataset, then evaluate metrics, deploy the model to an endpoint, and run online and batch predictions.

Create a text entity extraction dataset in Vertex AI via the GCP console, configure data type and objective, and upload or import text files to prepare for training.

Train a text entity extraction model in Vertex AI AutoML using the text entity extraction dataset, set a model name and id, and start training; then view in Model Registry.

Deploy the trained text entity extraction model by creating an endpoint in Vertex AI, deploying to US Central1, and testing predictions for disease entities with high confidence.

Use Google Colab to perform online and batch text entity extraction predictions with Vertex AI, installing and configuring GCP packages, authenticating, and linking to a dataset, model, and endpoint.

Use an endpoint for online predictions with predict, retrieving confidences and text segment offsets, and prepare JSON line inputs to run batch predictions in Vertex AI, review outputs.

Learn machine learning for video data to classify, detect, track, and recognize actions, and generate video content. Build models with CNNs on frames, horizon based CNNs, RNNs, and 3D CNNs.

Leverage AutoML to create models for video data, analyzing and extracting information from video content, and train models for action recognition, video classification, and object tracking.

Explore auto ML for video action recognition and how models identify actions in video sequences. Apply to surveillance, elderly monitoring, HCI, video retrieval, and summarization, with walking, bending, and falling.

Train a video action recognition model with AutoML on Vertex AI, starting from a Vertex AI dataset, then deploy an endpoint and run online and batch predictions.

Create a dataset from the MIT gold swing recognition subset stored in a public cloud bucket, using a CSV with video uri, time segment start and end, and swing annotation.

Create a Vertex AI video action recognition dataset named Golf Swing, with 100 videos, 70 training and 30 testing, using csv inputs from Cloud Storage.

Train a video action recognition model with AutoML in Vertex AI using the golf swings dataset. Set data split and review precision 0.92 and recall 90%.

Create and run a batch prediction workflow for video action recognition on Google Cloud using a Json lines batch input file and Vertex AI to detect goal swing.

Define tabular data as a table with rows and columns, like Excel spreadsheets or databases. Differentiate directly measurable features from calculated features, and define feature engineering as deriving new features.

Learn how Vertex AI handles tabular data with classification (binary and multi-class), regression, and forecasting. Use features and a designated target column to train multiple models by changing the target.

Explore time series fundamentals, including trend, seasonal, cyclic, and random patterns, and learn how forecasting uses historical data to predict future values for stock, energy, sales, weather, and web traffic.

Build and train a time series forecasting model with Vertex AI AutoML on a managed CSV dataset, using budget and optimization objectives for batch predictions on Iowa liquor sales.

Explore time series forecasting methods by contrasting qualitative and quantitative approaches, including scenarios with and without historical data. Learn about statistical methods and ML models used in time series analysis.

Explore the context window and forecasting horizon influence on time series forecasts, including past data points, window size and position, and how they shape models' ability to capture trends and seasonality.

Build a time series dataset from the 2020 Iowa liquor sales data. Use date as the time index and store name as the id; the target is sales dollars.

Create a tabular forecasting dataset in Vertex AI, upload a CSV, set store name as the series identifier and date as timestamp, then train an AutoML model with daily granularity.

Train a time series forecasting model for Iowa liquor sales in Vertex AI, then run a batch prediction on a CSV in cloud storage to forecast 30 days of sales.

Explore image classification fundamentals and build a dogs-versus-cats classifier with Google's AutoML, using no-code or minimal-code options, and understand single-label versus multi-label classification.

Prepare training data to build your auto ML based image classification dataset, train the model, evaluate with testing data, iterate, and deploy to production for predictions.

Prepare training data by creating a dataset from the dogs versus cats collection, a publicly available set of about 25,000 images, with 2,000 used for training an image classification model.

Enable Vertex AI and cloud storage APIs, create a cloud bucket, and build a Vertex AI image classification single label dataset; upload and label cat and dog images for training.

Train an image classification model with AutoML on a dog versus cat dataset, evaluate with a confusion matrix and 100% precision/recall, then deploy an endpoint and test with uploaded images.