Research map / Detection and segmentation / Open-set and few-shot detection
Real-Time Object Detection and Skeletonization for Motion Prediction in Video Streaming
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's full text. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Presents SkelAI, a full-stack app combining YOLOv8 segmentation, OpenCV contour extraction, and medial axis skeletonization for real-time skeletonization of video streams, with DALL-E 3 synthetic images and PyTorch-compatible skeleton export.
Problem
Real-time object detection and medial axis skeletonization for live video streams, without requiring pre-segmented masked frames, to support motion prediction and sequence-model training.
Approach
Chains YOLOv8 instance segmentation with OpenCV contour extraction, Gaussian noise reduction, and medial axis computation to skeletonize detected objects on arbitrary backgrounds in video streams, and adds DALL-E 3 generated images as extra inputs and export of skeleton data for training sequence models.
Key results
- Average compression ratio of 39:1 across five test images (original image size vs. medial axis size, Table 1).
- Claims improved accuracy and noise resistance over existing tools, but no quantitative comparison is reported in the text.
Limitations
- Evaluation is minimal: only a compression ratio on five images; no accuracy, latency, or baseline comparison reported.
- Claimed improvements over existing tools are not quantified.
- Sports injury and autonomous car applications are speculative and untested.
Experimental setup
Metrics: compression ratio
Baselines: previous Skeleton App (Kumar et al. 2024)
Builds on: YOLOv8 OpenCV
Closest papers by meaning
- MoVie: Broaden Your Views with Human Motion for Action Detection (CVPR 2026 · similarity 0.85)
- DeepFlux for Skeletons in the Wild (CVPR 2019 · similarity 0.84)
- Human Motion Prediction via Learning Local Structure Representations and Temporal Dependencies (AAAI 2019 · similarity 0.84)
- Spatial Temporal Graph Convolutional Networks for Skeleton-Based Action Recognition (AAAI 2018 · similarity 0.84)
- Spatio-Temporal Graph Convolution for Skeleton Based Action Recognition (AAAI 2018 · similarity 0.84)
- Revealing Key Details to See Differences: A Novel Prototypical Perspective for Skeleton-based Action Recognition (CVPR 2025 · similarity 0.83)
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.