Research map / Detection and segmentation / CNN features
ElasticFormer: Detecting Objects in HRW Shots via Elastic Computing Vision Transformer
Wenxi Li, Jingchen Huang, Chenyang Lyu, Moran Liu, Haozhe Lin, Guiguang Ding, Yuchen Guo
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 |
|---|---|---|---|
| replace fixed-ratio SparseFormer backbone with ElasticFormer | DINO · PANDA | AP50: 0.780 -> 0.806 | Improved |
| replace SparseFormer with ElasticFormer | DINO · PANDA | overall GFLOPs: 75.71 -> 13.13 | Improved |
| replace SparseFormer with ElasticFormer | Dynamic-Head · PANDA | AP50: 0.771 -> 0.782 | Improved |
| replace SparseFormer with ElasticFormer | Modified-DINO · PASCAL VOC 2007 (WSOD) | mAP: 50.3 -> 50.9 | Improved |
| use 3-phase training instead of 1-phase (ablation) | ElasticFormer (pseudo-GT) · PANDA | AP50: 0.784 -> 0.806 | Improved |
| supervise with pseudo-GT instead of GT (ablation) | ElasticFormer (3-phase) · PANDA | AP50: 0.798 -> 0.806 | Improved |
| increase gamma from 0.1 to 0.3 (ablation) | ElasticFormer · PANDA | AP50 / GFLOPs: 0.806 -> 0.795 AP50; 13.13 -> 10.28 GFLOPs | Mixed |
| lower alpha from 1.0 to 0.8 (ablation) | ElasticFormer · PANDA | AP50: 0.802 -> 0.806 | Improved |
| lower alpha from 0.8 to 0.6 (ablation) | ElasticFormer · PANDA | AP50: 0.806 -> 0.799 | Hurt |
| use variance scoring instead of small learned objectness (ablation) | ElasticFormer · PANDA | AP50 / latency: 0.803 -> 0.806 AP50; 1.213 -> 0.397 ms | Mixed |
| use variance scoring instead of gradient density (ablation) | ElasticFormer · PANDA | AP50: 0.798 -> 0.806 | Improved |
| use ElasticNet (conv) instead of ResNet-50 backbone | DINO · PANDA | AP50 / GFLOPs: 0.542 -> 0.754 AP50; 124.24 -> 16.54 GFLOPs | Improved |
Closest papers by meaning
- Elastic ViTs from Pretrained Models without Retraining (NeurIPS 2025 · similarity 0.84)
- ElasticDiffusion: Training-free Arbitrary Size Image Generation through Global-Local Content Separation (CVPR 2024 · similarity 0.83)
- MatFormer: Nested Transformer for Elastic Inference (NeurIPS 2024 · similarity 0.83)
- FSHNet: Fully Sparse Hybrid Network for 3D Object Detection (CVPR 2025 · similarity 0.83)
- SparseFormer: Sparse Visual Recognition via Limited Latent Tokens (ICLR 2024 · similarity 0.82)
- ElasticTok: Adaptive Tokenization for Image and Video (ICLR 2025 · similarity 0.82)
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.
