IDEA-Research / DINO

[ICLR 2023] Official implementation of the paper "DINO: DETR with Improved DeNoising Anchor Boxes for End-to-End Object Detection"
Apache License 2.0
2.1k stars 230 forks source link

Is there any way to get the coordinates(polygon shape returns)? #171

Closed yellowjs0304 closed 1 year ago

yellowjs0304 commented 1 year ago

Hi, First of all Thank you for sharing nice work. I'm beginner of DETR, and your model looks better than DETR.

I didn't check the detailed results yet, but i think this model also following DETR's output like bounding box (x, y, w, h). Is there any way to get the polygon shaped bbox? (before : x, y, w, h -> after : x1, y1, x2, y2, x3, y3, x4, y4).

My dataset has not only rotated but also, distorted. I can also apply Segmentation task, but segmentation accuracy is really low in my data, so I can't get the contour of that segmentation mask.

If you have any ideas, please help me :)

HaoZhang534 commented 1 year ago

Do you mean you want to predict distorted and rotated boxes? Our model does not support til now.

yellowjs0304 commented 1 year ago

I got it. Thanks for reply.