30 credits – Really, really fast tracking in image space

Scania is now undergoing a transformation from being a supplier of trucks, buses and engines to a supplier of complete and sustainable transport solutions.

Background
Autonomous vehicle development at Scania is advancing at a very high pace and self-driving trucks and buses on public roads will soon see the light of day. Autonomous Transport Solutions (ATS) Research at Scania is responsible for developing, testing and piloting future frontier ATS concepts. This work is done using agile and self-steered teams with the ambition to detect and evaluate upcoming technologies and prepare these for industrialization. We work in close cooperation with Volkswagen Group Innovation, leading technology suppliers and academic institutions.

This thesis work will lie under the supervision of the AI team, called EARA in Scania lingua. The group EARA develops deep learning-based methods that are used in the scene perception.

You will work closely with the members of this highly competent multicultural team, instrumental in developing cutting-edge autonomous technologies where your ideas will be encouraged and embraced.

Description
Tracking in image space is a well-known topic in computer vision, which has many applications. One such application is in automotive industry as input to later parts of a perception chain. An example application consuming such tracks could be tracking in 3d space with inputs from multiple sensor, not only cameras.

For tracking in image space to be useful in an automotive application it has to be accurate, fast and light on resources. This is typically a pick one or two out of three problem, you cannot have all of it at the same time.

Tracking by detection has been the most popular state of the art approach to tracking in image space the last decade. It is built around the idea that a really good object detection algorithm outputs high quality detections. These are fed into a tracking algorithm that outputs tracks. This problem is typically split into a GPU part doing the object detections via a deep neural network, and a CPU part that uses these detections over time to form tracks. One example of such an algorithm is the well-known Simple On-line and Real-time Tracking (SORT) which achieved good accuracy and a very fast run-time due to its relative simplicity. It was also extended to DeepSORT, where visual features extracted from the object detection model were utilized to improve accuracy.

The last couple of years, focus has shifted to also do tracking as part of a DNN either as an extension to an existing network or introducing a new tracking specific network. However, when introducing an extension or a new network, it has to compete with already existing networks for resources, e.g., semantic segmentation networks or the object detection network. Moreover, if a new network is introduced it also needs to be trained and have access to training data. There are, thus, some constraints to consider when trying to introduce a new network, and the more lightweight the network is, the more likely it will be that it is usable in a real setting, given that the accuracy is good enough. Lately some lightweight approaches have been considered in the literature, namely Tracktor and CenterNet, trying as SORT did 5 years ago: reducing the problem as much as possible and check what the real performance will be.

This thesis' problem formulation is thus:


• based on Tracktor, CenterNet and other recent similar work, what would be good candidates to a lightweight tracking DNN for Scania?
• implement one of these candidate DNNs and compare run-time, accuracy and resource utilization to Scania’s tracking algorithms.

The successful applicant will have the opportunity to gain hands-on experience of working on the latest sensors, computing platform and Scania’s concept autonomous vehicles. The applicant will also have access to the knowledgeable researchers and developers working at Scania’s Autonomous Transport Solutions Pre-Development & Research department.

We are open to the exploration of innovative ideas and if feasible the applicant might also get a chance to submit her/his results to a reputable research conference or even submit a patent application.

Applicants
One thesis worker studying a master's program in Computer Science, Electrical Engineering or similar. Applicants are expected to have a good understanding of computer vision, machine learning and practice thereof. The applicant should have sufficient software development knowledge to be able to implement/analyse mathematical concepts. Prior experience with deep learning is a plus. The applicant should be able to work in a diverse environment and communicate effectively in English. The personal traits of being agile, giving/receiving constructive feedback and taking initiatives will come handy.

Time plan
The project is planned for 20 weeks and can be started any time in early Spring 2022.
Applicants will be assessed on continuous basis until the position is filled.

Communication of the results
The results will be described in a report, published on Scania’s internal web and by the applicants’ university, and be shown in presentations at Scania. The prototype tool will be made available to Scania.

Organisation
The project will be performed within Scania’s Autonomous Transport Solutions Pre-Development & Research department.

Contacts
Thomas Gustafsson, PhD, Expert Engineer, AI Technologies, Autonomous Transport Solutions, thomas.gustafsson@scania.com

 


Scania is a world-leading provider of transport solutions. Together with our partners and customers we are driving the shift towards a sustainable transport system. In 2020, we delivered 66,900 trucks, 5,200 buses as well as 11,000 industrial and marine power systems to our customers. Net sales totalled to over SEK 125 billion, of which over 20 percent were services-related. Founded in 1891, Scania now operates in more than 100 countries and employs some 50,000 people. Research and development are mainly concentrated in Sweden. Production takes place in Europe and Latin America with regional product centres in Africa, Asia and Eurasia. Scania is part of TRATON GROUP. For more information visit: www.scania.com.