Skip to content

MarcelVSHNS/StixelNExT_Pro

Repository files navigation

StixelNExT++

📢 INFO: This repository will be fully published upon the official publication announcement.


🚀 Sneak Preview

StixelNExT++ introduces a novel approach to monocular scene representation, extending the Stixel World paradigm. Our method infers 3D Stixels and enhances object segmentation by clustering smaller Stixel units, creating a highly compressed yet flexible representation for point clouds and bird’s-eye-view (BEV) maps.

Key Features

Real-time performance – Achieves inference speeds as fast as 10 ms per frame.
Lightweight neural network – Trained with automatically generated LiDAR-based ground truth - holistic and specific.
Strong performance – Evaluated on the Waymo dataset, delivering competitive results within a **30-meter range **.
Versatile applications – Ideal for collective perception in autonomous systems.


📊 Results

Our model was evaluated on the Waymo dataset, focusing on vehicles and pedestrians. Image features were extracted and projected into metric space using camera intrinsics. Stixels are represented by their start and end points in Cartesian space.

Waymo Dataset – Stixel Prediction

Result 3D Result

By clustering Stixels, we achieve competitive object detection performance.

Stixel Clustering for Object Detection

Clustering Result

We also trained models on the KITTI dataset, expanding our approach to capture all possible obstacles, including buildings, vegetation, and environmental structures.

KITTI Dataset – Holistic Scene Representation

KITTI Result


🔜 Stay Tuned!

More updates coming soon! 🚀

About

3D Stixel World from AI

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages