- Applications are considered on a rolling basis
- Stockholm
- On-site
Job Description
Research Context
This internship is embedded within an active PhD research project aimed at benchmarking State-of-the-Art (SOTA) Multi-Object Tracking (MOT) solutions specifically for Human-Robot Interaction (HRI).
In HRI maintaining a consistent identity for users is critical. Unlike autonomous driving or surveillance Furhat observe the world from a stationary eye-level perspective which presents unique challenges. The core research investigates how different inputs (e.g. Face vs. Full-Body bounding boxes) and detection qualities impact the robots ability to maintain a social interaction.
Internship Topic
The exact topic is not set in stone; we will define it together.
We are looking for a student to take ownership of a specific part of the pipeline. Depending on your interests and strengths your internship could for example focus on one or a combination of the following:
- Benchmarking SOTA Trackers: You could focus on the evaluation side running specific trackers (like ByteTrack OC-SORT Bot-SORT ) on our custom dataset to measure their performance in HRI scenarios.
- Data Collection: You could design and execute new experiments to capture videos of natural Human-Robot Interaction expanding the diversity of our dataset.
- Fine-Tuning & Annotation: You could help annotate data to create Ground Truths that allow us to fine-tune trackers for our needs.
Regardless of the specific direction the internship will involve close collaboration with the PhD supervisor to ensure the work contributes to the wider research paper.
Required Experience:
Intern
Applications are considered on a rolling basis StockholmOn-siteJob DescriptionResearch ContextThis internship is embedded within an active PhD research project aimed at benchmarking State-of-the-Art (SOTA) Multi-Object Tracking (MOT) solutions specifically for Human-Ro...
- Applications are considered on a rolling basis
- Stockholm
- On-site
Job Description
Research Context
This internship is embedded within an active PhD research project aimed at benchmarking State-of-the-Art (SOTA) Multi-Object Tracking (MOT) solutions specifically for Human-Robot Interaction (HRI).
In HRI maintaining a consistent identity for users is critical. Unlike autonomous driving or surveillance Furhat observe the world from a stationary eye-level perspective which presents unique challenges. The core research investigates how different inputs (e.g. Face vs. Full-Body bounding boxes) and detection qualities impact the robots ability to maintain a social interaction.
Internship Topic
The exact topic is not set in stone; we will define it together.
We are looking for a student to take ownership of a specific part of the pipeline. Depending on your interests and strengths your internship could for example focus on one or a combination of the following:
- Benchmarking SOTA Trackers: You could focus on the evaluation side running specific trackers (like ByteTrack OC-SORT Bot-SORT ) on our custom dataset to measure their performance in HRI scenarios.
- Data Collection: You could design and execute new experiments to capture videos of natural Human-Robot Interaction expanding the diversity of our dataset.
- Fine-Tuning & Annotation: You could help annotate data to create Ground Truths that allow us to fine-tune trackers for our needs.
Regardless of the specific direction the internship will involve close collaboration with the PhD supervisor to ensure the work contributes to the wider research paper.
Required Experience:
Intern
View more
View less