Glmv Grounding

A skill that uses GLM-V native grounding capabilities for coordinate conversion, bounding-box visualization, and more. GLM-V native grounding can locate any target specified by the prompt in an image and output relative coordinates normalized to 0-1000 based on image size. Coordinate formats include 2D bounding box (default), 2D points, and 3D bounding box. GLM-V also supports spatiotemporal localization and tracking of multiple prompt-specified targets in videos, outputting 2D bounding boxes per second.

zai-org Updated

File contents

zai-org/GLM-skills/tree/main/skills/glmv-grounding commit 21ec30aa5a

Frequently asked questions

npx skillmds@latest add zai-org/glmv-grounding-2