# Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity

Source: [Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity](<https://www.sota2.com/research/paper/68fa6bf61e209131cea31475>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2025-10-28T13:56:52.201Z

Authors: Byungseok Roh, JaeWoong Shin, Wuhyun Shin, Saehoon Kim

Published: 2021-11-29T05:22:46.000Z

## Abstract

DETR is the first end-to-end object detector using a transformer encoder-decoder architecture and demonstrates competitive performance but low computational efficiency on high resolution feature maps. The subsequent work, Deformable DETR, enhances the efficiency of DETR by replacing dense attention with deformable attention, which achieves 10x faster convergence and improved performance. Deformable DETR uses the multiscale feature to ameliorate performance, however, the number of encoder tokens increases by 20x compared to DETR, and the computation cost of the encoder attention remains a bottleneck. In our preliminary experiment, we observe that the detection performance hardly deteriorates even if only a part of the encoder token is updated. Inspired by this observation, we propose Sparse DETR that selectively updates only the tokens expected to be referenced by the decoder, thus help the model effectively detect objects. In addition, we show that applying an auxiliary detection loss on the selected tokens in the encoder improves the performance while minimizing computational overhead. We validate that Sparse DETR achieves better performance than Deformable DETR even with only 10% encoder tokens on the COCO dataset. Albeit only the encoder tokens are sparsified, the total computation cost decreases by 38% and the frames per second (FPS) increases by 42% compared to Deformable DETR.   Code is available at https://github.com/kakaobrain/sparse-detr

## Recorded benchmark results

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

- [Object Detection on VisDrone (val)](<https://www.sota2.com/research/sota/object-detection-on-visdrone-val.md>): Object Detection; VisDrone (val); AP50: 42.5.

- [Object Detection on COCO 2017 (val)](<https://www.sota2.com/research/sota/object-detection-on-coco-2017-val.md>): Object Detection; COCO 2017 (val); AP: 46.3.

## Original sources and implementation

[Original paper](<https://arxiv.org/abs/2111.14330>)

[PDF](<https://arxiv.org/pdf/2111.14330.pdf>)
