mmdetection-to-tensorrt icon indicating copy to clipboard operation
mmdetection-to-tensorrt copied to clipboard
verified publisher icon grimoire

[Feature Request] GroundingDINO support

Open codingbus821 opened this issue 4 months ago • 0 comments

Request Can you add support for GroundingDINO model conversion to TensorRT? Why GroundingDINO?

Popular open-vocabulary object detection model Text + image input for flexible detection Good accuracy but slow inference in PyTorch Would benefit greatly from TensorRT optimization

Current Status

✅ Works in MMDetection ❌ Not supported in mmdetection-to-tensorrt

Technical Notes

Uses dual inputs (image + text tokens) Has BERT text encoder component Cross-modal attention layers

Question Are there any plans to support GroundingDINO? I'm interested in using this model with TensorRT optimization.

codingbus821 avatar Aug 11 '25 05:08 codingbus821