IDEA-Research / T-Rex

[ECCV2024] API code for T-Rex2: Towards Generic Object Detection via Text-Visual Prompt Synergy
https://deepdataspace.com/blog/T-Rex
Other
2.28k stars 147 forks source link

跑偏了。 #48

Closed xiyangyang99 closed 7 months ago

xiyangyang99 commented 7 months ago

groundingdino和本文类似,但是在工程运用上,不好落地,推理速度不像yolo-world那般快。transformer不好加速推理。跨模态方式的文本加载方式在加速上处理较为麻烦。

Mountchicken commented 7 months ago

Hi @xiyangyang99 Indeed, DETR-based models face significant deployment and speed challenges, and we are trying to optimize them. Yet from the perspective of open-set detection performance, both GroundingDINO and T-Rex deliver remarkably better results.

chooooock commented 4 weeks ago

groundingdino和本文类似,但是在工程运用上,不好落地,推理速度不像yolo-world那般快。transformer不好加速推理。跨模态方式的文本加载方式在加速上处理较为麻烦。

在部署Gdino到trt框架的时候遇到问题,在很多project的issues上看到你的回到,方面分享一下Gdino到onnx再到trt的可行方法吗?或者留个联系方式可以吗?