RF-Signal-based Human Pose Estimation
- Did camera calibration and RF device embedded software to construct a synchronous collection system to collect a dataset of image and RF signal with more than 10,000 RF signals and corresponding images.
- Deployed OpenPose and AlphaPose for automatic annotation of optical images; developed the visual annotation tool to refine the results of the automatic annotation.
- Trained the network model based on Cross-Model Supervision method using PyTorch; reached super resolution estimation result on test set; succeeded to do pose estimation through curtain without any training.
