detrex: Benchmarking Detection Transformers

• DOI: 10.48550/arxiv.2306.07265
• Publication Date: 2023-01-01
• Authors: Tianhe Ren; Shilong Liu; Feng Li; Hao Zhang; Ailing Zeng; Jie Yang; Xingyu Liao; Ding Jia; Hongyang Li; He Cao; Jianan Wang; Zhaoyang Zeng; Xianbiao Qi; Yuhui Yuan; Jianwei Yang; Lei Zhang
• Keywords: Benchmarking; Benchmark (surveying); Codebase

The DEtection TRansformer (DETR) algorithm has received considerable attention in the research community and is gradually emerging as a mainstream approach for object detection other perception tasks. However, current field lacks unified comprehensive benchmark specifically tailored DETR-based models. To address this issue, we develop unified, highly modular, lightweight codebase called detrex, which supports majority of instance recognition algorithms, covering various fundamental tasks, including detection, segmentation, pose estimation. We conduct extensive experiments under detrex perform Moreover, enhance performance transformers through refinement training hyper-parameters, providing strong baselines supported algorithms.We hope that could offer communities standardized platform to evaluate compare different models while fostering deeper understanding driving advancements recognition. Our code available at https://github.com/IDEA-Research/detrex. project currently being actively developed. encourage use further development contributions.