IMUTube: Automatic extraction of virtual on-body accelerometry from video for human activity ...
summary
IMUTube converts online video into virtual on-body accelerometers by applying 2D and 3D pose estimation and signal processing to reconstruct body motion and compute accelerometry at specific body sites (example: right knee of a cyclist). Benchmarks show historic wearable datasets have only ~20+ participants and Opportunity contains >70% Null labels, while IMUTube-trained random-forest models produce F1-scores comparable to real-data models, mixing virtual+real boosts performance, and transfer learning handles missing labels.
AI-written summary · about
related talks
06mApproaching the Real-World: Supporting Activity Recognition Training with Virtual IMU Data
13mIMUCoCo: Enabling Flexible On-Body IMU Placement for Human Pose Estimation and Activity Recognition
13mUltraPoser: Pushing the Limits of IMU-based Full-Body Pose Estimation with Ultrasound Sensing on ...
11mBaroPoser: Real-time Human Motion Tracking from IMUs and Barometers in Everyday Devices
22mRunning, walking, sitting or biking? - Motion prediction with acceleration and rotation
24mProject GaitNet: Ushering in the ImageNet moment for human Gait kinematics
23mCounting exercise repetition using live camera video
28mDeep Learning-Based Human Activity Recognition for Physical Skills...