Such a good explanation! Thank you :) Why do the authors choose a fixed number of denoised boxes instead of using the number of actual ground truth boxes? So what happens if there are e.g. 15 objects in the ground truth and there are only 10 denoising queries?
Great question! Total number of denoising queries is gt_objects * number_of_groups. So for 5 groups, 15 gt objects - you will have 75 denoising queries. For 2 gt objects, you will have just 10 denoising queries
Most certainly will! I am doing DINO after SWIN, which will be my next video. It's also one of the DINO components. You can expect DINO in around 3 weeks. Rt detr will take a bit more time, but I will definitely do it!
After this video i will try Dn detr
Happy to hear that
I was trying Dab detr in the morning
Keep exploring the topic! Hopefully my videos could help
Keep going 💪
Thanks! I will!
Finally, bro ❤🎉
))))
I realy support you 😊
Warms my heart!
Check out my next video: Reading DN Detr source code th-cam.com/video/6DDJjoLtsyA/w-d-xo.html
Such a good explanation! Thank you :)
Why do the authors choose a fixed number of denoised boxes instead of using the number of actual ground truth boxes?
So what happens if there are e.g. 15 objects in the ground truth and there are only 10 denoising queries?
Great question!
Total number of denoising queries is gt_objects * number_of_groups. So for 5 groups, 15 gt objects - you will have 75 denoising queries. For 2 gt objects, you will have just 10 denoising queries
@@makgaiduk ah i didnt get that, thanks!
I am making a next video about reading the source code for dn detr. Helps clear out ambiguities like these )@@davidro00
@@makgaiduk looking forward💪🏼
@@makgaiduk Hi! Thank you for a great video! Can you explain what "group" means here?
would you please review rt-detr and DINO detr?
Most certainly will! I am doing DINO after SWIN, which will be my next video. It's also one of the DINO components. You can expect DINO in around 3 weeks. Rt detr will take a bit more time, but I will definitely do it!