Recommended size of fine-tuning dataset

lyuwenyu / RT-DETR

[CVPR 2024] Official RT-DETR (RTDETR paddle pytorch), Real-Time DEtection TRansformer, DETRs Beat YOLOs on Real-time Object Detection. 🔥 🔥 🔥

Apache License 2.0

2.31k stars 258 forks source link

Recommended size of fine-tuning dataset #408

Open retazo0018 opened 1 month ago

retazo0018 commented 1 month ago

Hi,

Thanks for the great work!

I would like to use RT-DETR for detecting novel objects. I would like to know the recommended number of instances per novel object category needed for a good detection accuracy?

Many Thanks,

VimukthiRandika1997 commented 1 month ago

It is always good to have a larger dataset: around 5000 images at least. This depends on the task you are working