Research map / Detection and segmentation / CNN features
DM-EFS: Dynamically Multiplexed Expanded Features Set Form for Robust and Efficient Small Object Detection
Aashish Sharma
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 EFS form (use higher-resolution shallow backbone features) (ablation) | YOLOv7 · VisDrone val (640x640) | AP50: 47.81 -> 52.01 | Improved |
| add EFS form (ablation) | YOLOv7 · VisDrone val (640x640) | fps: 39.71 -> 32.14 | Hurt |
| add dynamic feature multiplexing (DFM) to EFS form (ablation) | YOLOv7 + EFS · VisDrone val (640x640) | fps: 32.14 -> 38.24 | Improved |
| add dynamic feature multiplexing (DFM) to EFS form (ablation) | YOLOv7 + EFS · VisDrone val (640x640) | AP50: 52.01 -> 51.80 | No clear effect |
| swap in size-features codebook learned on DarkFace (ablation) | DM-EFS · VisDrone val (640x640) | AP: 29.71 -> 29.13 | No clear effect |
| swap in size-features codebook learned on VisDrone (ablation) | DM-EFS · DarkFace val (640x640) | AP: 29.30 -> 28.95 | No clear effect |
| raise lambda_szp to 0.04 from <=0.02 (ablation) | DM-EFS · VisDrone val (640x640) | AP50 | Improved |
| raise lambda_szp to >=0.06 (ablation) | DM-EFS · VisDrone val (640x640) | AP50 | Hurt |
| use DM-EFS instead of ESOD with same YOLOv7 base | YOLOv7 · VisDrone val (640x640) | AP: 27.16 -> 29.71 | Improved |
| use DM-EFS instead of ESOD with RetinaNet base | RetinaNet · VisDrone val (640x640) | AP: 08.91 -> 09.11 | Improved |
| use DM-EFS instead of ESOD with YOLOv5 base | YOLOv5 · VisDrone val (640x640) | AP: 23.61 -> 25.14 | Improved |
| add DM-EFS to YOLOv7 on SODA-D | YOLOv7 · SODA-D test (640x640) | AP: 30.52 -> 34.91 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- YOLOv7 (Base) — YOLOv7: Trainable Bag-of-Freebies Sets New State-of-the-Art for Real-Time Object Detectors (CVPR 2023) · better on 11 of 13 shared results
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
- ScaleKD: Distilling Scale-Aware Knowledge in Small Object Detector (CVPR 2023 · similarity 0.86)
- Expanding Low-Density Latent Regions for Open-Set Object Detection (CVPR 2022 · similarity 0.86)
- Towards Large-Scale Small Object Detection: Survey and Benchmarks (TPAMI 2023 · similarity 0.86)
- Oriented Tiny Object Detection: A Dataset, Benchmark, and Dynamic Unbiased Learning (TPAMI 2025 · similarity 0.86)
- Reviving the Discarded High-Resolution Feature for Transformer-Based Tiny Object Detection (NeurIPS 2026 · similarity 0.85)
- DFD: Distilling the Feature Disparity Differently for Detectors (ICML 2024 · similarity 0.84)
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.
