Research map / Vision-language and multimodal / Cross-modal learning
BridgeNet: Comprehensive and Effective Feature Interactions via Bridge Feature for Multi-Task Dense Predictions
Jingdong Zhang, Jiayuan Fan, Peng Ye, Bo Zhang, Hancheng Ye, Baopu Li, Yancheng Cai, Tao Chen
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 Bridge Feature Extractor (BFE) to MTL baseline (ablation) | MTL baseline · NYUD-v2 (ViT-base) | Delta_MTL (%): -1.17 -> 0.96 | Improved |
| add Task Pattern Propagation (TPP) on top of BFE (ablation) | BFE · NYUD-v2 (ViT-base) | Delta_MTL (%): 0.96 -> 1.29 | Improved |
| add Task Feature Refiner TFR-huge on top of BFE+TPP (ablation) | BFE+TPP · NYUD-v2 (ViT-base) | Delta_MTL (%): 1.29 -> 2.45 | Improved |
| use TFR-base instead of pixel-wise addition for refinement (ablation) | bridge-feature refinement · NYUD-v2 (ViT-base) | Delta_MTL (%): 0.96 -> 1.52 | Improved |
| use TFR-base instead of InvPT for refinement, with half the parameters (ablation) | bridge-feature refinement · NYUD-v2 (ViT-base) | Delta_MTL (%): 1.52% > 0.97% | Improved |
| scale TFR from base to large to huge (ablation) | TFR · NYUD-v2 (ViT-base) | Delta_MTL (%): 1.52 -> 1.96 -> 2.45 | Improved |
| concatenate bridge and task features with one conv instead of TFR (ablation) | bridge-feature refinement · NYUD-v2 (ViT-base) | Delta_MTL (%): 2.45 -> 1.16 | Hurt |
| train Seg. and Dep. jointly (ablation) | BridgeNet TFR-base · NYUD-v2 task sets | Delta_MTL (%): +5.70% | Improved |
| train Seg. and Norm. jointly (ablation) | BridgeNet TFR-base · NYUD-v2 task sets | Delta_MTL (%): -0.51% | Hurt |
| train Dep. and Norm. jointly (ablation) | BridgeNet TFR-base · NYUD-v2 task sets | Delta_MTL (%): +2.94% | Improved |
| train Seg. and H.Parts jointly (ablation) | BridgeNet TFR-base · PASCAL Context task sets | Delta_MTL (%): 1.20% | Improved |
| add Norm. to the Seg., H.Parts and Sal. task set (ablation) | BridgeNet TFR-base · PASCAL Context task sets | Delta_MTL (%): -0.33% -> -2.70% | Hurt |
Closest papers by meaning
- A Dynamic Feature Interaction Framework for Multi-task Visual Perception (IJCV 2023 · similarity 0.87)
- TaskPrompter: Spatial-Channel Multi-Task Prompting for Dense Scene Understanding (ICLR 2023 · similarity 0.87)
- Going Beyond Multi-Task Dense Prediction with Synergy Embedding Models (CVPR 2024 · similarity 0.87)
- TaskExpert: Dynamically Assembling Multi-Task Representations with Memorial Mixture-of-Experts (ICCV 2023 · similarity 0.86)
- VMT-Adapter: Parameter-Efficient Transfer Learning for Multi-Task Dense Scene Understanding (AAAI 2024 · similarity 0.85)
- UVU: Improving Multimodal Understanding via Vision-Language Unified Autoregressive Paradigm (CVPR 2026 · 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.
