Research map / Detection and segmentation / CNN features
Rethinking Learnable Tree Filter for Generic Feature Transform
Lin Song, Yanwei Li, Zhengkai Jiang, Zeming Li, Xiangyu Zhang, Hongbin Sun, Jian Sun, Nanning Zheng
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 LTF-V2 at stages 345 | Mask R-CNN with ResNet-50 · COCO 2017 val | AP_box: 38.8 -> 41.2 | Improved |
| Add LTF-V2 at stages 345 | Mask R-CNN with ResNet-50 · COCO 2017 val | AP_seg: 35.2 -> 37.0 | Improved |
| Add LTF-V2 | ResNet-101 segmentation model · Cityscapes test | mIoU: 82.1 | Improved |
| Add unary and pairwise terms together (ablation) | LTF-V2 on ResNet-50 · COCO 2017 val | AP_box: 39.7 -> 40.7 | Improved |
| Add the unary term to pairwise term and LST (ablation) | LTF-V2 on ResNet-50 · COCO 2017 val | AP_box: 40.2 -> 41.2 | Improved |
| Add the pairwise term to unary and pairwise-free variant (ablation) | LTF-V2 on ResNet-50 · COCO 2017 val | AP_box: 39.7 -> 40.7 | Improved |
| Add the learnable spanning tree (ablation) | LTF-V2 on ResNet-50 with ImageNet pretraining · COCO 2017 val | AP_box: 40.9 -> 41.2 | Improved |
| Replace MST with LST (ablation) | LTF-V2 on ResNet-50 trained from scratch · COCO 2017 val | AP_box: 29.4 -> 30.3 | Improved |
| Add LTF-V2 (ablation) | ResNet-50 segmentation model · Cityscapes val, no test-time augmentation | mIoU: 72.9 -> 77.4 | Improved |
| Add LTF-V2 (ablation) | ResNet-50 segmentation model · Cityscapes val, multi-scale and flip | mIoU: 75.5 -> 78.9 | Improved |
| Insert LTF-V2 at stages 345 rather than stage 4 only (ablation) | Mask R-CNN with ResNet-50 · COCO 2017 val | AP_box: 40.6 -> 41.2 | Improved |
| Use 16 groups instead of 1 group (ablation) | LTF-V2 on Mask R-CNN with ResNet-50 · COCO 2017 val | AP_box: 40.2 -> 41.2 | Improved |
Closest papers by meaning
- Learnable Tree Filter for Structure-preserving Feature Transform (NeurIPS 2019 · similarity 0.95)
- Tree Energy Loss: Towards Sparsely Annotated Semantic Segmentation (CVPR 2022 · similarity 0.86)
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.85)
- Hierarchy-Agnostic Unsupervised Segmentation: Parsing Semantic Image Structure (NeurIPS 2024 · similarity 0.85)
- Bridge the Points: Graph-based Few-shot Segment Anything Semantically (NeurIPS 2024 · similarity 0.84)
- Unlocking the Potential of Pre-trained Vision Transformers for Few-Shot Semantic Segmentation through Relationship Descriptors (CVPR 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.
