Pith. sign in

REVIEW 8 cited by

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

Not yet reviewed by Pith; the record is open.

Re-run · record.json Download PDF Read on arXiv ↗

This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.

SPECIMEN: schema-true, not a live event

T0 review · schema-true

One-sentence machine reading of the paper's core claim.

pith:XXXXXXXX · record.json · timestamp

arxiv 2111.14330 v2 pith:U4MG6U44 submitted 2021-11-29 cs.CV cs.LG

Byungseok Roh , JaeWoong Shin , Wuhyun Shin , Saehoon Kim This is my paper · ORCID
classification cs.CVcs.LG
keywords detrencoderperformancedeformabletokensonlyattentiondetection
verification ladder T0 review T1 audit T2 compute T3 formal

Signed reviews

No signed human review yet.

0 comments
read the original 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

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 8 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. OpenAlex reports about 91 citations worldwide. Full citation record

  1. Focus Through Motion: RGB-Event Collaborative Token Sparsification for Efficient Object Detection

    cs.CV 2025-09 conditional novelty 6.0 of 10

    FocusMamba uses event-camera activity to adaptively prune uninformative tokens in both RGB and event streams, improving detection accuracy and cutting FLOPs.

  2. MTVNet: Mapping using Transformers for Volumes -- Network for Super-Resolution with Long-Range Interactions

    cs.CV 2024-12 conditional novelty 6.0 of 10

    MTVNet is a multi-scale transformer for volumetric super-resolution that uses carrier tokens to carry information across large 3D regions, setting new benchmarks on high-resolution bone CT data.

  3. DS-Det: Single-Query Paradigm and Attention Disentangled Learning for Flexible Object Detection

    cs.CV 2025-07 conditional novelty 5.0 of 10

    A flexible-query transformer detector that separates cross-attention localization from self-attention deduplication reports consistent accuracy gains over DINO across five backbones.

  4. Decoupled PROB: Decoupled Query Initialization Tasks and Objectness-Class Learning for Open World Object Detection

    cs.CV 2025-07 conditional novelty 5.0 of 10

    Decoupled PROB combines query selection with learnable queries and stops objectness prediction early in the decoder, improving known-object detection in open-world benchmarks.

  5. Trexplorer Super: Topologically Correct Centerline Tree Tracking of Tubular Objects in CT Volumes

    cs.CV 2025-07 conditional novelty 5.0 of 10

    Trexplorer Super improves centerline tree tracking in CT volumes using Super Trajectory Training, Focal Cross Attention, and Target Augmentation, and outperforms prior models on new synthetic and real datasets.

  6. Object Detection using Event Camera: A MoE Heat Conduction based Detector and A New Benchmark Dataset

    cs.CV 2024-12 conditional novelty 5.0 of 10

    MvHeat-DET, a Mixture-of-Experts heat-conduction detector, outperforms 15 re-trained baselines on the new EvDET200K dataset and reaches 55.7 mAP on N-Caltech101.

  7. Underwater object detection in sonar imagery with detection transformer and Zero-shot neural architecture search

    cs.CV 2025-05 reject novelty 4.0 of 10

    NAS-DETR uses a maximum-entropy zero-shot neural architecture search to build a CNN-Transformer backbone for a Deformable-DETR detector, reporting improved sonar object detection on URPC2021 and URPC2022.

  8. X-ray illicit object detection using hybrid CNN-transformer neural network architectures

    cs.CV 2025-05 conditional novelty 4.0 of 10

    On multi-scanner X-ray data, hybrid CNN-transformer detectors outperform the CNN-only YOLOv8 baseline, while the CNN baseline remains best on single-scanner benchmarks.

Pith tools