DETR

DETR (DEtection TRansformer) is a novel object detection framework that utilizes a pure transformer encoder-decoder architecture to directly predict object bounding boxes and classes without the need for hand-designed components like Non-Maximum Suppression (NMS) or anchor generation. / DETR是一种新颖的目标检测框架，利用纯Transformer编码器-解码器架构直接预测目标边界框和类别，无需手动设计的组件，如非极大值抑制（NMS）或锚框生成。

Knowledge type

数学理论

First proposed or discovered

2020

Proposer

Carlo R. Carion

Creator

农场主

Last updated

2026-10-08 07:59:29