[Computer Vision] Visual SLAM - Honda Research Institute USA

[Computer Vision] Visual SLAM

[Computer Vision] Visual SLAM

Job Number: P20INT-47
This position focuses on using monocular camera combined with other sensors (e.g. GPS, Odometry) to conduct point cloud map creation, developing efficient map data structure with special emphasis on handling everyday appearance changes and long term environmental effects on the visual map. The candidate is expected to develop robust software for real vehicles using a combination of well-established and novel research techniques.
San Jose, CA

You are expected to: 

  • Survey state of the art in Long Term Localization, Map Management and Visual Place Recognition.

  • Propose an architecture for handling everyday appearence and long term environmental changes.

  • Prepare datasets for Long Term Localization.

  • Implement the proposed architecture and integrate it into the existing framework.

Minimum Qualifications:

  • Ph.D. or highly qualified M.S. candidate in computer science, electrical engineering, or related field.

  • Prior research and development experience in SLAM is required.

  • Must be familiar with the Visual SLAM pipeline and able to write working code.

  • Hands-on experience in one or more of the following: sensor fusion, pose graph optimization, 3D reconstruction, long-term localization.

  • Excellent programming skills in C++ and Python.

Bonus Qualifications:

  • Familiarity with Deep Learning Frameworks such as TensorFlow, PyTorch.
     
  • Familiarity with Learning-Based Visual Place Recognition architectures such as NetVLAD is a plus.
     
  • Familiarity with Learning-Based Feature Extractors such as SuperPoint is a plus.

Duration: 3 months

How to apply

Candidates must have the legal right to work in the U.S.A.​ Please add Cover Letter and CV in the same document