search

IDEA-Research / GroundingDINO

[ECCV 2024] Official implementation of the paper "Grounding DINO: Marrying DINO with Grounded Pre-Training for Open-Set Object Detection"
https://arxiv.org/abs/2303.05499
Apache License 2.0
6.2k stars 652 forks source link

Testing or evaluating code in COCO dataset #51

Open ml-ai-cv opened 1 year ago

ml-ai-cv commented 1 year ago

Hello,

I am trying to replicate the performance of Grounding DINO on COCO dataset. Is there any code or any way to replicate the performance of the Grounding DINO on COCO dataset?

Thanks for the great work!

Dwrety commented 1 year ago

I wrote a piece of code to test on COCO 2017val. And I got an AP around 40.8. I am also writing the training code myself, but basically right now the code of "Language-guide Query Selection" is not included. So I guess the performance is deteriorated because of this.

kobic8 commented 1 year ago

I wrote a piece of code to test on COCO 2017val. And I got an AP around 40.8. I am also writing the training code myself, but basically right now the code of "Language-guide Query Selection" is not included. So I guess the performance is deteriorated because of this.

hey, I'm also intrested in fine-tuning this model to a custom labeled dataset I have, are you planning to share your traning code?

mary-0830 commented 1 year ago

I wrote a piece of code to test on COCO 2017val. And I got an AP around 40.8. I am also writing the training code myself, but basically right now the code of "Language-guide Query Selection" is not included. So I guess the performance is deteriorated because of this.

Hello, thank you for sharing. May I ask if you could share your evaluation code?