Research map / Detection and segmentation / CNN features
Efficient Decoder-Free Object Detection with Transformers
Peixian Chen, Mengdan Zhang, Yunhang Shen, Kekai Sheng, Yuting Gao, Xing Sun, Ke Li, Chunhua Shen
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 |
|---|---|---|---|
| use the DOT backbone instead of Swin (ablation) | DFFT-Small · COCO DFFT-Small 12 epochs | AP (%): 33.8 -> 37.9 | Improved |
| add the scale-aggregated encoder (SAE) (ablation) | DFFT-Small with DOT · COCO DFFT-Small 12 epochs | AP (%): 37.9 -> 39.9 | Improved |
| add the task-aligned encoder (TAE) (ablation) | DFFT-Small with DOT and SAE · COCO DFFT-Small 12 epochs | AP (%): 39.9 -> 41.4 | Improved |
| add TAE to DOT without SAE (ablation) | DFFT-Small with DOT · COCO DFFT-Small 12 epochs | AP (%): 37.9 -> 39.8 | Improved |
| replace SW-MSA with global channel-wise attention (GCA) (ablation) | DOT backbone · COCO DFFT-Small 12 epochs | AP (%): 39.0 -> 40.1 | Improved |
| add one extra GCA block (ablation) | DOT backbone · COCO DFFT-Small 12 epochs | AP (%): 40.1 -> 40.4 | Improved |
| replace the extra GCA with semantic-augmented attention (SAA) (ablation) | DOT backbone · COCO DFFT-Small 12 epochs | AP (%): 40.4 -> 41.4 | Improved |
| use SAA instead of FPN (ablation) | RetinaNet head on DFFT backbone · COCO 12 epochs | AP (%): 38.4 -> 38.9 | Improved |
| use SAE instead of YOLOF dilated encoder (ablation) | DFFT-Small · COCO DFFT-Small 12 epochs | AP (%): 40.3 -> 41.4 | Improved |
| use SAE instead of concatenation (ablation) | DFFT-Small · COCO DFFT-Small 12 epochs | AP (%): 39.6 -> 41.4 | Improved |
| scale up from Nano to Medium | DFFT · COCO 36 epochs | AP (%): 42.8 -> 45.7 | Improved |
| replace decoder-based DETR with DFFT-Nano | DETR-R50 · COCO | AP (%): 42.0 -> 42.8 | Improved |
Closest papers by meaning
- End-to-End Object Detection with Transformers (ECCV 2020 · similarity 0.94)
- ViDT: An Efficient and Effective Fully Transformer-based Object Detector (ICLR 2022 · similarity 0.93)
- WB-DETR: Transformer-Based Detector Without Backbone (ICCV 2021 · similarity 0.92)
- Sparse DETR: Efficient End-to-End Object Detection with Learnable Sparsity (ICLR 2022 · similarity 0.91)
- DECO: Unleashing the Potential of ConvNets for Query-based Detection and Segmentation (ICLR 2025 · similarity 0.90)
- DETRs Beat YOLOs on Real-time Object Detection (CVPR 2024 · similarity 0.89)
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.
