Giters

IDEA-Research / MaskDINO

[CVPR 2023] Official implementation of the paper "Mask DINO: Towards A Unified Transformer-based Framework for Object Detection and Segmentation"

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Questions about segmentation head

sunlight002 opened this issue · comments

commented

image
Why does Cb used the 1/4 resolution feature map from the backbone instead of using the multi scale features? If use the multi scale features it will hava a better performer for small object?

commented

Hi, we did not investigate this problem. Maybe multi-scale features will be helpful.

commented

Thanks for your reply !