Downloads · 30 days
0
microduck-pose-gesture
XenderYang/microduck-pose-gesture
microduck-pose-gesture is a machine learning model from XenderYang. Use it for the machine learning task on the model card, and read the license before you ship it in a product.
Head-cam person pose/gesture model for the duck robot (sim-first, sim2real-ready). - Detects person + keypoints (COCO-17 subset) from duck first-person RGB (320x240 train / 640x480 deploy), IMX219-like 62° HFOV, 40° h…
Access
Public
Updated Sep 8, 2026
Repo size
83.9 MB
Likes
0
Public
Repo makeup
Click a slice to open those files.
.onnx83.9 MB · 100%
At a glance
- Access
- Public
- Created
- Sep 8, 2026
- Updated
- Sep 8, 2026
- SHA
- 39afc4bb
Model card
From the Hugging Face model README
MicroDuck Pose-Gesture (YOLO11m-pose, v8)
Head-cam person pose/gesture model for the duck robot (sim-first, sim2real-ready).
- Detects person + keypoints (COCO-17 subset) from duck first-person RGB (320x240 train / 640x480 deploy), IMX219-like 62° HFOV, 40° head pitch (dynamic), 25cm eye height.
- Gestures (2D-separable at low view): idle / right_fwd / left_fwd / both_fwd / both_side.
- Dataset: synthetic v8, renderer-exact keypoint labels + domain randomization (floor/brightness/noise/blur/JPEG/jitter), balanced.
- Validation: Box mAP50 ≈ 0.995, Pose mAP50 ≈ 0.85 (150-val).
- Contract: ONNX
[1,3,480,480] -> [1,56,4725]; decode yolov8/yolo11 pose. See https://github.com/YDxun/microduck-duck-play for training/benchmark code.
- Created
- Sep 8, 2026
- Updated
- Sep 8, 2026