Research map / Detection and segmentation / CNN features
Slim-DETR: Real-Time Tiny Object Detection with Efficient Interaction and Gaussian Query
Tong Wu, Jiahao Zhang, Juntao Guan, Lai Rui
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| add multi-scale Transformer encoder layer (Variant A) (ablation) | RT-DETR baseline encoder · VisDrone val (B2-B4 input) | AP: 30.7 -> 31.3 | Improved |
| add multi-scale Transformer encoder layer (Variant A) (ablation) | RT-DETR baseline encoder · VisDrone val (B2-B4 input) | latency (ms): 100 -> 166.66 | Hurt |
| add gather-and-inject mechanism (Variant B over A) (ablation) | encoder with multi-scale layer · VisDrone val (B2-B4 input) | AP: 31.3 -> 33.1 | Improved |
| add gather-and-inject mechanism (Variant B over A) (ablation) | encoder with multi-scale layer · VisDrone val (B2-B4 input) | latency (ms): 166.66 -> 28.57 | Improved |
| add FSDA foreground-token attention (Variant C over B) (ablation) | EI encoder · VisDrone val (B2-B4 input) | AP: 33.1 -> 33.3 | Improved |
| add FSDA foreground-token attention (Variant C over B) (ablation) | EI encoder · VisDrone val (B2-B4 input) | latency (ms): 28.57 -> 24.39 | Improved |
| add gather-and-inject mechanism (Variant B over A) (ablation) | encoder with multi-scale layer · VisDrone val (B3-B5 input) | AP: 29.1 -> 30.0 | Improved |
| add FSDA (Variant C over B) (ablation) | EI encoder · VisDrone val (B3-B5 input) | AP: 30.0 -> 30.3 | Improved |
| use higher-resolution B2 features instead of B3-B5 (full variant C) (ablation) | EI encoder · VisDrone val | AP: 30.3 -> 33.3 | Improved |
| add CBAM and gate to gathered features (ablation) | EI encoder · VisDrone val | feature map quality | Improved |
| replace uncertainty-minimal query selection with GTG (6 layers) (ablation) | Slim-DETR-R50 · VisDrone val | AP: 33.3 -> 33.5 | Improved |
| reduce decoder depth from 6 to 4 layers with GTG (ablation) | Slim-DETR-R50 · VisDrone val | AP: 33.5 -> 33.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- DQ-DETR — DQ-DETR: Dual Query Detection Transformer for Phrase Extraction and Grounding (AAAI 2023) · better on 3 of 3 shared results
- Salience-DETR — Salience DETR: Enhancing Detection Transformer with Hierarchical Salience Filtering Refinement (CVPR 2024) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity (ICLR 2022 · similarity 0.88)
- DETRs Beat YOLOs on Real-time Object Detection (CVPR 2024 · similarity 0.85)
- MI-DETR: An Object Detection Model with Multi-time Inquiries Mechanism (CVPR 2025 · similarity 0.85)
- Sparse Semi-DETR: Sparse Learnable Queries for Semi-Supervised Object Detection (CVPR 2024 · similarity 0.85)
- Learning 1-Bit Tiny Object Detector with Discriminative Feature Refinement (ICML 2024 · similarity 0.85)
- State Space Model Meets Transformer: A New Paradigm for 3D Object Detection (ICLR 2025 · similarity 0.85)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
