Research map / Vision-language and multimodal / Cross-modal learning
DF2Net: Discriminative Feature Learning and Fusion Network for RGB-D Indoor Scene Classification
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 structured (triplet-style) embedding loss to RGB stream (ablation) | AlexNet RGB · SUN RGB-D | mean class accuracy (%): 42.6% -> 46.3% | Helped |
| add structured embedding loss to Depth stream (ablation) | AlexNet Depth · SUN RGB-D | mean class accuracy (%): 38.4% -> 39.2% | Helped |
| initialize fusion streams with DFL-trained single-modal networks (ablation) | two-stream RGB+D concat · SUN RGB-D | mean class accuracy (%): 48.3% -> 51.7% | Helped |
| add structured loss on the concatenated fc7 feature (ablation) | DFL for RGB+D · SUN RGB-D | mean class accuracy (%): 51.7% -> 51.7% | No clear effect |
| add correlative embedding on holistic RGB and Depth features (ablation) | DFL for RGB+D · SUN RGB-D | mean class accuracy (%): 51.7% -> 52.9% | Helped |
| split features into distinctive and correlative parts instead of fully correlating (ablation) | DFL for RGB+D & CFL · SUN RGB-D | mean class accuracy (%): 52.9% -> 54.6% | Helped |
| add structured embedding loss to RGB stream (ablation) | AlexNet RGB · NYU Depth V2 | mean class accuracy (%): 59.5% -> 61.1% | Helped |
| add structured embedding loss to Depth stream (ablation) | AlexNet Depth · NYU Depth V2 | mean class accuracy (%): 49.3% -> 54.8% | Helped |
| initialize fusion streams with DFL-trained networks (ablation) | two-stream RGB+D concat · NYU Depth V2 | mean class accuracy (%): 60.6% -> 63.5% | Helped |
| add distinctive and correlative fusion (DCFL) (ablation) | DFL RGB+D · NYU Depth V2 | mean class accuracy (%): 63.5% -> 65.4% | Helped |
Closest papers by meaning
- Discriminative Multi-Modal Feature Fusion for RGBD Indoor Scene Recognition (CVPR 2016 · similarity 0.92)
- Locality-Sensitive Deconvolution Networks With Gated Fusion for RGB-D Indoor Semantic Segmentation (CVPR 2017 · similarity 0.87)
- Depth CNNs for RGB-D Scene Recognition: Learning from Scratch Better than Transferring from RGB-CNNs (AAAI 2017 · similarity 0.86)
- Modality and Component Aware Feature Fusion For RGB-D Scene Classification (CVPR 2016 · similarity 0.85)
- H2O-SDF: Two-phase Learning for 3D Indoor Reconstruction using Object Surface Fields (ICLR 2024 · similarity 0.83)
- Multi-Modal Object Re-identification via Sparse Mixture-of-Experts (ICML 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.