MediaPipe processed pose CSVs
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
This upload contains MediaPipe Pose Landmarker 3D landmark outputs as CSV files, used for bias and error evaluation against Unity ground-truth joint trajectories in a controlled multi-camera grid setup. Archive: mediapipe_processed_csv.zip (~30.3 MB). MD5: e10454e2f88c05aeb09531f53a71c36c Layout after extraction (match the analysis repository under 02_mediapipe_processed/): mediapipe_processed_csv/ Y=0.5/ CapturedFrames___.csv Y=1.0/ Y=1.5/ Y=2.0/ Each filename encodes camera position as CapturedFrames_{X}_{Y}_{Z} where the middle value is camera height in metres (0.5, 1.0, 1.5, or 2.0). CSV columns: frame_id, landmark, x, y, z, visibility, image_path (MediaPipe normalised / model space as produced by the batch pipeline). These files are the intermediate layer between raw capture images and downstream statistics (joint-angle MAE, direction-angle analysis, etc.). Raw images and Unity synced_joint_positions.csv are not included in this archive; obtain GT and images separately if full reproduction is required. Code and documentation: see the related GitHub repository (analysis pipeline, run.py, verify_paper_data.py, docs/REPRODUCTION.md).
Fields of Study & Subjects
Human Pose and Action Recognition
Hand Gesture Recognition Systems
Human Motion and Animation
Publication details
- Journal / source
- Zenodo (CERN European Organization for Nuclear Research)
- Publication date
- 28 March 2026
- Language
- EN
- Document type
- Dataset