Detrex is a research platform for DETR-based visual recognition tasks including object detection, instance segmentation, and pose estimation. It provides implementations of state-of-the-art detection transformers such as DINO, Deformable DETR, Mask DINO, and DAB-DETR, along with training pipelines and benchmarks. The library is built on PyTorch and aims to unify and accelerate research on transformer-based detection models.