sparolab

sparolab / KISS-IMU

Public

KISS-IMU: Self-supervised Inertial Odometry with Motion-balanced Learning and Uncertainty-aware Inference. @ ICRA'26 Award Finalist

49
2
100% credibility
Found May 09, 2026 at 49 stars -- GitGems finds repos before they trend. Get early access to the next one.
Sign Up Free
AI Analysis
Python
AI Summary

KISS-IMU is a research codebase for training neural networks to correct raw IMU sensor data into accurate motion estimates using self-supervision from LiDAR scans.

How It Works

1
🔍 Discover KISS-IMU

You stumble upon this robotics breakthrough that makes motion tracking from phone-like sensors super reliable, even in tough spots.

2
📥 Grab the files

Download everything and sort your sensor readings and real paths into neat folders, just like the guide shows.

3
⚙️ Quick setup

Choose the easy all-in-one launcher or simple install to get your workspace ready in moments.

4
📊 Feed in your data

Connect your collection of movement recordings so the system learns from balanced examples of walking, turning, and more.

5
🚀 Train the motion tracker

Hit start to teach it how to clean up shaky sensor data using smart self-learning tricks.

6
🧪 Test on new paths

Run checks on fresh sequences to measure how accurately it follows positions and turns.

🏆 Pinpoint motion mastery

Celebrate rock-solid tracking results with plots and scores rivaling top research, ready for your robot adventures.

Sign up to see the full architecture

5 more

Sign Up Free

Star Growth

See how this repo grew from 49 to 49 stars Sign Up Free
Repurpose This Repo

Repurpose is a Pro feature

Generate ready-to-use prompts for X threads, LinkedIn posts, blog posts, YouTube scripts, and more -- with full repo context baked in.

Unlock Repurpose
AI-Generated Review

What is KISS-IMU?

KISS-IMU is a Python library for self-supervised inertial odometry that denoises raw IMU streams using LiDAR-odometry pseudo-labels, delivering motion-balanced learning and uncertainty-aware inference. It tackles IMU drift in robotics by training neural corrections via simple scripts like train.sh and evaluate.sh, supporting datasets such as KITTI, MulRan, and DiTer-OS out of the box. Developers get sliding-window evaluation metrics like RPE and APE, with Docker for instant setup.

Why is it gaining traction?

As an ICRA'26 award finalist, KISS-IMU stands out with its GMM sampler and frequency gating to prevent dominant motions from overwhelming training, paired with KISS-ICP IMU backends for robust loop closure. The quick-start Docker mounts your data and runs training or baselines like raw PVGO in one command, while tunable env vars make hyperparameter sweeps straightforward. It pairs IMU preintegration with adaptive covariances, yielding tighter trajectories than vanilla methods.

Who should use this?

Robotics engineers building legged robots or ground vehicles needing drift-free IMU odometry in GPS-denied environments. SLAM researchers experimenting with imuno kiss or imusic kiss fusions on forest/night sequences. Devs prototyping uncertainty-aware inference for real-time inertial navigation.

Verdict

Promising for IMU odometry research thanks to the ICRA'26 finalist nod and polished Docker/scripts, but low 49 stars and 1.0% credibility score signal early-stage maturity—expect tweaks for production. Grab it if you're tinkering with self-supervised learning on LiDAR-IMU datasets.

(198 words)

Sign up to read the full AI review Sign Up Free

Similar repos coming soon.