Gemini Enterprise Agent Platform Vision release notes

Model output visualization tool in vaictl

You can now visualize model output using the vaictl command line tool available through the Vertex AI Vision SDK. This visualization combines real-time video streams with annotations and statistics generated by Vertex AI Vision models. This feature lets you better understand model output by visualizing output information over your video stream.

Features:

• Visualize command added to vaictl command-line tool.
• Support added to visualize Occupancy Analytics model annotations.

For more information, see Visualize model output.

Motion filtering - motion detection zone

You can now specify zones for motion detection, or areas to omit from motion detection. For more information, see the motion filtering guide.

Batch video and image support in Vertex AI Vision Warehouse is Generally Available. Vertex AI Vision Warehouse now supports semantic searches and similarity searches on video and images. For more information, see Vision Warehouse overview

Image batch processing now available as a Preview feature

Vertex AI Vision now offers batch image processing as a Preview feature. This new processing mode lets you provide a Cloud Storage path with image files as input and Cloud Storage path to store output batch processing results.

For more information, see the image batch processing documentation.

Python SDK now available

A new Python SDK is now available for Vertex AI Vision. For more information, see the following documentation pages:

• Set up a project and a development environment
• Use the Face Blur model with the Python SDK