The solution depicted in the architecture diagram on the right deploys a pipeline that is triggered when you add a new PDF document to your Cloud Storage bucket. The pipeline extracts text from your document, creates a summary from the extracted text, and stores the summary in a database for you to view and search.

You can invoke the application by either uploading files via Jupyter Notebook, or directly to Cloud Storage in the Google Cloud console.

The solution, depicted in the diagram on the right, uses pretrained machine learning models to analyze images provided by users and generate image annotations. Deploying this solution creates an image processing service that can help you handle unsafe or harmful user-generated content, digitize text from physical documents, detect and classify objects in images, and more.

You will be able to review configuration and security settings to understand how to adapt the image processing service to different needs.

Before analyzing your video data with your application, create a pipeline for the continuous flow of data with Streams service in Vertex AI Vision. Ingested data is then analyzed by Google’s pretrained models or your custom model. The analysis output from the streams is then stored in Vertex AI Vision Warehouse where you can use advanced AI-powered search capabilities to query unstructured media content.

Visual Inspection AI is optimized in every step so it’s easy to set up and fast to see ROI. With up to 300 times fewer labeled images to start training high-performance inspection models than general purpose ML platforms, it has shown to deliver up to 10 times higher accuracy. You can train models with no technical expertise, and they run on-premises. Best of all, the models can be continuously refreshed with data flowing from the factory floor, giving you increased accuracy as you discover new use cases.