30 Credit - Multi-frame detection and tracking with offline perception networks

Introduction:
This thesis will be carried out at the Autonomous Solutions department at Scania, where we develop cutting-edge research solutions for scene perception, prediction and planning for autonomous driving. 

Thesis work is an excellent way to get closer to Scania and build relationships for the future. Many of our colleagues began their Scania career with their degree project.

Background:
Autonomous driving relies on a precise understanding of the surrounding environment, achieved through the analysis of sensor data by perception algorithms. A key component of this process is the ability to identify and localize objects of interest (detection) and to capture their dynamics by following them over time (tracking). In recent years, deep neural networks have demonstrated substantial improvements over traditional computer vision approaches in both tasks.

Perception carried out in real time on the vehicle is referred to as online perception. These algorithms must function under strict resource constraints, particularly regarding latency and memory.

In contrast, offline perception is performed on powerful servers outside the vehicle.  Offline models are widely used for data curation and in auto-labelling pipelines. Such pipelines exploit the outputs of highly accurate offline perception models as priors, which can then be refined and processed into reliable labels. This enables the scalable generation of large, high-quality datasets essential for training robust online perception models.

Objective:
The objective of this thesis is to investigate and develop methods for integrated detection-and-tracking networks with camera and/or LiDAR input. Unlike standard online perception, where networks must operate in real-time on board a vehicle, this work takes place offline. This enables the use of multiple frames of sensor data, both from the past and the future, providing information unavailable in real-time settings and thereby improving the accuracy and robustness of the generated tracks. 

Job Description:
The work will include, but not be limited to, the following tasks:

•    Conduct a literature review on state-of-the-art methods in detection, and multi-frame tracking for autonomous driving.
•    Develop methods for bounding box detection with integrated tracking and explore how access to future context can improve detection and tracking of dynamic actors. A promising approach is described in [3].
•    Design approaches to evaluate and ensure the quality of resulting dynamic actor trajectories

The successful candidate will have the opportunity to work with rich multimodal datasets, cutting-edge perception architectures, and the latest research challenges in autonomous transport. Collaboration with experienced researchers and developers in Scania’s Autonomous Transport Solutions Pre-Development & Research department will provide both industrial relevance and academic depth to the work.

Qualifications:
-    Currently enrolled in a Master program in Computer Science, Electrical Engineering or related field
-    Good understanding of computer vison, machine learning and ML frameworks such as PyTorch
-    Proficiency in programming languages such as Python and/or C++
-    Able to work in a diverse environment and communicate effectively in English
-    Excellent problem-solving skills and the ability to work independently

Time plan: 
The project is planned for 20 weeks and can be started any time in early Spring 2025. Applicants will be assessed on a continuous basis until the position is filled.

Number of students: 1-2

Contact persons and supervisors:
Caroline Heidenreich, caroline.heidenreich@scania.com, Development Engineer, Autonomous Transport Solutions, Traton AB
Giulia D’Ascenzi, giulia.dascenzi@scania.com, Development Engineer, Autonomous Transport Solutions, Traton AB

Application:
Your application must include a CV, personal letter and transcript of grades.

A background check might be conducted for this position. We are conducting interviews continuously and may close the recruitment earlier than the date specified.