Research map / Detection and segmentation / Open-set and few-shot detection
Time-rEversed diffusioN tEnsor Transformer: A New TENET of Few-Shot Object Detection
Shan Zhang, Naila Murray, Lei Wang, Piotr Koniusz
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 third- and fourth-order tensors to second-order (r=2,3,4 vs r=2) (ablation) | TENET · VOC2007 novel, 5/10-shot | mAP: outperforms r=2 by 5.8% and 2.7% | Improved |
| use higher orders in RPN (settings a vs b) (ablation) | TENET · VOC2007 novel 5/10-shot | mAP: 53.4 -> 57.2 | Improved |
| use higher orders in both RPN and TRH (a vs c) (ablation) | TENET · VOC2007 novel 5/10-shot | mAP: 53.4 -> 61.0 | Improved |
| add first-order features alongside higher orders in TRH (c vs d) (ablation) | TENET · VOC2007 novel 5-shot | mAP: 61.0 -> 62.3 | Improved |
| add Tensor Shrinkage Operator (eta>1 vs eta=1) (ablation) | TENET · VOC2007 novel | mAP: ~ 5% gain | Improved |
| add TSO (eta 7,7,7 vs 1,1,1) (ablation) | TENET · FSOD dataset 5-shot | AP50: 30.8 -> 35.4 | Improved |
| add Spatial-HOP head to Z-shot head (ablation) | TENET TRH · VOC2007 novel 5-shot | mAP: 58.5 -> 62.3 | Improved |
| add Z-shot head to Spatial-HOP head (ablation) | TENET TRH · VOC2007 novel 5-shot | mAP: 61.0 -> 62.3 | Improved |
| channel split 5:2:1 instead of 6:1:1 (ablation) | TENET · VOC2007 novel 5-shot | mAP: 53.6 -> 62.3 | Improved |
| number of attention heads TA 4 vs 1 (ablation) | TENET · VOC2007 novel 5-shot | mAP: 58.3 -> 61.8 | Improved |
| increase TENET blocks TB beyond 2 (ablation) | TENET · VOC2007 novel 5-shot | mAP: 62.3 -> 62.1 | No clear effect |
| Swin-B7 backbone instead of ResNet-50 (ablation) | TENET · VOC2007 novel 5/10-shot | mAP: 62.3 -> 62.6 | No clear effect |
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.
- DeFRCN — DeFRCN: Decoupled Faster R-CNN for Few-Shot Object Detection (ICCV 2021) · better on 2 of 2 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
- Learning General and Specific Embedding with Transformer for Few-Shot Object Detection (IJCV 2024 · similarity 0.88)
- Temporal Object-Aware Vision Transformer for Few-Shot Video Object Detection (AAAI 2026 · similarity 0.88)
- MSTDiff: Multiscale-Aware Transformer Diffusion Network for Video Object Detection (AAAI 2026 · similarity 0.88)
- Deeply Tensor Compressed Transformers for End-to-End Object Detection (AAAI 2022 · similarity 0.87)
- TGBFormer: Transformer-GraphFormer Blender Network for Video Object Detection (AAAI 2025 · similarity 0.86)
- Multi-Perspective Data Augmentation for Few-shot 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.
