Research map / 3D vision / Human pose estimation
3D Human Sensing, Action and Emotion Recognition in Robot Assisted Therapy of Children With Autism
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 |
|---|---|---|---|
| Fine-tune DMHS for partially visible people (ablation) | DMHS pose and body-part labeling · H80KPartial | semantic body-part labeling accuracy: 59.6% -> 78.0% | Helped |
| Fine-tune DMHS for partially visible people (ablation) | DMHS pose estimation · H80KPartial | mean per joint position error: 79.6 mm -> 57.6 mm | Helped |
| Fine-tune DMHS for partially visible people (ablation) | DMHS pose and body-part labeling · Human80K | semantic body-part labeling accuracy: 79.0% -> 79.9% | Helped |
| Fine-tune DMHS for partially visible people (ablation) | DMHS pose estimation · Human80K | mean per joint position error: 63.3 mm -> 63.9 mm | Hurt |
| Include therapist pose features (ablation) | Moving Pose action classifier · DE-ENIGMA action classification | accuracy: 46.96% -> 47.49%; 32.92% -> 34.95%; 40.83% -> 44.14; 43.53% -> 45.07%; 44.20% -> 45.68% | Helped |
| Use acceleration components in the pose descriptor (ablation) | Moving Pose action classifier · DE-ENIGMA action classification | accuracy | No clear effect |
| Use a CNN instead of Moving Pose | Action classification using Kinect 3D pose features · DE-ENIGMA action classification | accuracy: 47.49% -> 53.1% | Helped |
| Use a CNN instead of Moving Pose | Action classification using DMHS-SMPL-T · DE-ENIGMA action classification | accuracy: 45.68% -> 47.9% | Helped |
| Use RGB DMHS-SMPL-T skeletons instead of Kinect skeletons | CNN emotion predictor · Valence prediction | RMSE: 0.116 -> 0.099 | Helped |
| Use RGB DMHS-SMPL-T skeletons instead of Kinect skeletons | CNN emotion predictor · Valence prediction | PCC: 0.184 -> 0.169 | Hurt |
| Use RGB DMHS-SMPL-T skeletons instead of Kinect skeletons | CNN emotion predictor · Arousal prediction | PCC: 0.345 -> 0.388 | Helped |
Closest papers by meaning
- $E^3$: Exploring Embodied Emotion Through A Large-Scale Egocentric Video Dataset (NeurIPS 2024 · similarity 0.84)
- Harmony4D: A Video Dataset for In-The-Wild Close Human Interactions (NeurIPS 2024 · similarity 0.82)
- Understanding Unique Behavioral Patterns through Multimodal Analysis of Eye-Hand Coordination in Autistic Children (Student Abstract) (AAAI 2025 · similarity 0.82)
- Weakly-Supervised Emotion Transition Learning for Diverse 3D Co-speech Gesture Generation (CVPR 2024 · similarity 0.81)
- Facial-R1: Aligning Reasoning and Recognition for Facial Emotion Analysis (AAAI 2026 · similarity 0.81)
- VidBot: Learning Generalizable 3D Actions from In-the-Wild 2D Human Videos for Zero-Shot Robotic Manipulation (CVPR 2025 · similarity 0.81)
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.