stera-sdk 0.0.4

SDK for processing data from Stera's mobile app: MCAP loading, visualization, and annotation

 Stera SDK

by FPV Labs (fpvlabs.ai)

stera-sdk is the Python SDK behind Stera (website · iOS app).

Load MCAP recordings, run swappable hand, face, and skeleton models behind one schema, and visualize or export clean episodes, ready for embodied AI, VLAs, and world models.

A short taste

import stera
from stera.data import MCAPReader
from stera.models import HandTracker, FaceBlurrer
from stera.viz import Visualizer
from stera.processing import MeshRefiner
from stera.eval import Evaluate

session = MCAPReader("recording.mcap")
hands   = HandTracker(model="mediapipe")
blur    = FaceBlurrer(model="mediapipe")
viz     = Visualizer(session, map_3d="auto")

for frame in session.frames():
    blurred = blur.blur(frame)
    poses   = hands.detect_hands(frame)
    session.add_rgb_frame(frame.index, blurred)
    session.add_hand_pose(frame.index, poses)
    viz.log_frame(frame, hands=poses)

mesh = MeshRefiner(session).refine()           # clean + densify + colorize
session.export("episodes/run_01", visualizer=viz, mesh=mesh)
Evaluate(session).show()                       # interactive HTML QC report

Getting started

pip install "stera-sdk[all]"

That pulls the core SDK plus MediaPipe, RetinaFace, and MeshRefiner deps. WiLoR, HaMeR, and EgoBlur ship as separate research repos, see the installation guide. ffmpeg must be on PATH for episode export.

Documentation

• 📚 Docs
• 🚀 Quickstart
• ⚙️ API reference
• 🛠️ Installation

License

Apache 2.0 © FPV Labs