Internship Offer – Automatic Segmentation of Domain Specific Vehicle in Point Clouds Using Superpoint Transformer

🏢 Company Overview

Perception4D is a leading engineering company specializing in 3D data analysis, with a strong focus on LiDAR-based perception.

We provide advanced development services, including sensor expertise, multi-sensor calibration and fusion, SLAM, object detection, and semantic segmentation. Our work supports applications in robotics, smart infrastructure, and collision detection. By combining state-of-the-art deep learning techniques with classical algorithms and mathematical modeling, we develop tailored approaches to complex perception challenges.

🚀 Internship Description

In this project, you will work with LiDAR data from a fixed sensor monitoring an area where vehicles are in motion. The objective is to detect, segment, and track unknown vehicles in the point clouds without prior knowledge of their shapes or models. The output will be used to automatically generate training data for a deep learning model that will later perform vehicle classification and recognition. To do this, you will perform semantic segmentation on the point clouds using Superpoint Transformer (https://arxiv.org/abs/2306.08045).

🎯 Your Mission:

• Understand and prepare raw LiDAR point clouds..
• Implement and fine-tune Superpoint Transformer for semantic segmentation.
• Process semantic data to filter out unwanted points (ground, buildings…)
• Develop tracking algorithms to follow detected vehicles across multiple frames.
• Build an automatic dataset generation pipeline to prepare for model specific training.
• Evaluate the system’s performance.

🧑‍🎓 Expected Qualifications and Mindset:

• Currently pursuing a Master’s degree in Engineering or Computer Science.
• Proficiency in the Python programming language with experience in machine learning frameworks and libraries (PyTorch, NumPy, etc.).
• (Bonus) Familiarity with GPU programming and CUDA.
• Knowledge of 3D point cloud data and LiDAR technology.
• Strong skills in linear algebra and geometry.
• Analytical and problem-solving mindset.
• Dynamic, eager to learn, autonomous, and proactive.

🧑‍🏫 The candidate will be supervised by an expert in the field, ensuring strong technical guidance throughout the internship.

Internship Details:

• Location: Lyon, France (Partial remote work possible)
• Duration: 6 months
• Start Date: Flexible

📨 If you are passionate about 3D vision and deep learning, apply now! Please submit a resume and cover letter (in English or French) to hiring@perception4d.com.