Thesis Work for Enabling Robot Vision: One-shot 6D Pose Estimation Benchmark for Machine Tending

Placering
Västerås

Thesis Work for Enabling Robot Vision: One-shot 6D Pose Estimation Benchmark for Machine Tending
At ABB, we are dedicated to addressing global challenges. Our core values: care, courage, curiosity, and collaboration - combined with a focus on diversity, inclusion, and equal opportunities - are key drivers in our aim to empower everyone to create sustainable solutions. That's our story. Make it your story.
This position reports to
R&D Team Lead

Your role and responsibilities
ABB Corporate Research Center in Sweden is located in Västerås, about 100 kilometers west of Stockholm. Together with our colleagues in other R&D centers we develop technologies for future products and services of ABB's core businesses. In the Robotics team we focus on robot concepts/ design, autonomous and intelligent robot, and robot applications in both industrial and service robot segments. Visual 6D pose estimation is crucial for robotics, especially for tasks such as grasping and manipulation. High-quality datasets with realism are essential for enabling robotic vision. Good benchmarks (including datasets and evaluation KPIs) allow us to find better models for practical use. Currently, there is a lack of datasets and benchmarks suitable for one-shot pose estimation of industry-relevant objects, especially for machine tending. This project targets for reviewing current 6D pose estimation models and datasets, and proposing a benchmark focusing on industry-relevant objects. Research topics: One-shot 6D pose estimation benchmarks for evaluating the robot vision, especially for machine tending cases. Goal(s): 1. Review of state-of-the-art 6D pose estimation methods and current available benchmarks (datasets and KPIs). 2. Propose benchmark, including: designing a method for developing and collecting a dataset for one-shot 6D pose estimation, defining different scenarios covering various cases for machine tending, and testing the state-of-the-art models on the collected dataset. Details: • 6 month (January, 2025 till June, 2025) • 30 credits • One student • Since lab work involved, the MS student need to come ABB office in Västerås • Contacts for details: [email protected] You will mainly be responsible for: • Review literature and IP on the topic of 6D pose estimation, datasets, and evaluation KPIs. Identify key metrics for one-shot estimation. • Propose benchmark for evaluating one-shot 6D pose estimation, specifically for machine tending. • Evaluate state-of-the-art model on proposed benchmark.

Qualifications for the role

• Suitable for student with interest in robotics and software engineering in general, and computer vision and deep learning in particular
• Good coding ability (Python required, C++ is merit)
• Deep learning and computer vision experience, such as object tracking, pose estimation is meriting

More about us
Recruiting Manager Liwei Qi, +46 73 021 23 09, and supervisor Chi Zhang, +46 72 461 26 88, will answer your questions. Positions are filled continuously. Apply with your CV, academic transcripts and a cover letter in English. We value people from different backgrounds. Could this be your story? Apply today or visit www.abb.com to read more about us and learn about the impact of our solutions across the globe. #MyABBStory #Li-onsite

Locations
Västerås, Sweden