Neural Radiance Fields (NeRFs) – Internship/Thesis

by

Website Mainblades

Mainblades Inspections is on a mission to change the way inspections are performed for large
commercial aircraft. Our company develops data-driven technology, focused on automation of the
complete inspection process: from autonomous drone flight to damage detection and assessment of
visual data, all the way to automated reporting. Our automated drone inspections for aircraft allow
for a more rapid and effective inspection at any location, enabling our customers to keep their aircraft
off the ground.

Research scope and objectives

The deep learning era began through the advancements it brought in traditional 2D imagerecognition tasks such as classifications, detections, and instance segmentations. As the techniques
matured, the research in deep-learning-based computer vision has been shifted towards
fundamental 3D computer vision problems — one of the most notable being synthesizing new views
of an object and reconstructing the 3D shape of it from images. A novel technique, Neural Radiance
Fields (NeRF), has quickly gained traction. A “vanilla” NeRF stores a volumetric scene representation
as the weights of an MLP, trained on many images with known poses. The effectiveness of such
method and its derivatives has led to more than 50 papers at CVPR 2022 related to NeRFs.

A key aspect of our product is to construct all individual photos into a single image or object to
deduplicate results which might be present in overlapping photos. To create a singular truth of the
results of an inspection, instead of relying on a multitude of photos. The goal of the assignment is to
explore the landscape of NeRF methods available and implement a suitable solution that allows us
to exploit all the sensor data that we collect with our autonomous drone to construct a realistic 3D
model based on the datasets that we have available from inspections. Relevant reading material:

  • Dense depth priors for NeRFs
  • Depth-supervised NeRF
  • Light Field Neural Rendering
  • Point-NeRF: Point-based Neural Radiance Fields
  • Neural Radiance Fields with Deterministic Integration for Volume Rendering

Requirements

What you bring to the table

We are looking for a talented and enthusiastic MSc student, preferably with some programming
experience and a background in computer vision and deep learning. Furthermore, we are looking for
people who like to be seriously challenged and easily adapt to change. You contribute innovative ideas
to the company and have a shared responsibility to make our vision a reality. We are looking for
someone with a dedication for mastering new topics and a willingness to learn.

What we bring to the table

A comfortable and fun working environment:

  • Collaborating with a young, dynamic, and multidisciplinary team in an innovative start-up.
  • Datasets and computational resources to conduct your research

If you think you got what it takes, and you are up for the challenge, take the first step and apply now!
If we are interested, we will reach out and take you through our recruitment process.

To apply for this job email your details to applications@mainblades.com