angular_app — Angular velocity / pose (Kivy)

Prototype Android-oriented / desktop Kivy app for joint angle analysis from video using a TensorFlow Lite pose model.

For physios and students exploring on-device pose angles (elbows, knees) without a full lab stack.

What it does

Single script: realtime_pose_kivy 4.py

• Loads a .tflite pose model and an .mp4 video via Kivy file choosers
• Estimates keypoints with TFLite Interpreter
• Computes angles for: prawy/lewy łokieć, prawe/lewe kolano
• Overlays pose and plots angles (OpenCV + Matplotlib + Kivy)

Requirements

• Python 3 with: kivy, opencv-python (cv2), numpy, matplotlib, tensorflow (TFLite Interpreter)

Quick start

git clone https://github.com/biokineticum/angular_app.git
cd angular_app
pip install kivy opencv-python numpy matplotlib tensorflow
python "realtime_pose_kivy 4.py"

Select a .tflite model and .mp4 video in the UI, choose a joint, then run analysis.

Note: filename contains a space — quote it on the command line as above.

---

Author

Dariusz Mosler (Biokineticum) — physiotherapy, biomechanics, sports science
🌐 biokineticum.com · 📰 Substack · 𝕏 @Biokineticum