What about results without coco pretrain?

Epiphqny / VisTR

[CVPR2021 Oral] End-to-End Video Instance Segmentation with Transformers

https://arxiv.org/abs/2011.14503

Apache License 2.0

740 stars 96 forks source link

Closed hehao13 closed 3 years ago

hehao13 commented 3 years ago

Hi! Thanks for your code. I wonder what are the results if we do not use coco pretrained detr ?

Epiphqny commented 3 years ago

hi @hehao13 , the results will be 8-10 points lower.