Research map / Vision-language and multimodal / Cross-modal learning
Multi-modal RGB–Depth–Thermal Human Body Segmentation
Cristina Palmero, Albert Clapés, Chris Bahnsen, Andreas Møgelmose, Thomas Baltzer Moeslund, Sérgio Escalera
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 |
|---|---|---|---|
| fuse color, depth, thermal and motion descriptions with a Random Forest (ablation) | best single description (HON) · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: 67.5 -> 78.6 | Improved |
| use learned (stacked) fusion instead of naive confidence averaging (ablation) | fusion of descriptions · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: 63.9 -> 78.6 | Improved |
| use Random Forest instead of MLP with gaussian activation for fusion (ablation) | learning-based fusion · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap improvement over HON: 5% -> 10% | Improved |
| use Random Forest instead of linear SVM for fusion (ablation) | learning-based fusion · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: +2.5% on average | Improved |
| use depth HON descriptor instead of motion HOF descriptor (ablation) | single-modality description · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: 14% better on average | Improved |
| use depth HON descriptor instead of color HOG (ablation) | single-modality description · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: 65 -> 67.5 | Improved |
| add all four modalities via RF fusion versus worst description (HOF) (ablation) | single descriptions · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: +25% on average | Improved |
| evaluate on thermal ground-truth masks instead of color/depth masks | RF fusion · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: relatively small decrease | Hurt |
| use RF fusion instead of Buys et al 2014 depth body-part method | segmentation method · RGB-D-T human body segmentation (leave-one-sequence-out) | overlap: 67.1 -> 78.6 | Improved |
| exclude sequence 10 from mean (poor-performing sequence) | RF fusion · RGB-D-T human body segmentation (leave-one-sequence-out) | mean overlap across sequences: almost 5% reduction due to Seq. 10 | Improved |
Closest papers by meaning
- A Focused Human Body Model for Accurate Anthropometric Measurements Extraction (CVPR 2025 · similarity 0.81)
- Human Activity Recognition using RGB-DVS Cameras: A Multi-modal Heat Conduction Model and A Benchmark Dataset (IJCV 2026 · similarity 0.81)
- Deep Multitask Architecture for Integrated 2D and 3D Human Sensing (CVPR 2017 · similarity 0.80)
- SHARP: Shape-Aware Reconstruction of People in Loose Clothing (IJCV 2022 · similarity 0.79)
- Multi-Modal Hand-to-Mouth Gesture Recognition in Activity-Oriented RGB-Thermal Footage (Student Abstract) (AAAI 2025 · similarity 0.77)
- Keep the Balance: A Parameter-Efficient Symmetrical Framework for RGB+X Semantic Segmentation (CVPR 2025 · similarity 0.77)
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.
