Next generation AI vision

You will be working on cutting-edge technologies that will pave the way for the next evolution in AI Vision.

The use of AI Vision (Artificial intelligence paired with traditional computer vision) for processing and making sense of 2-dimensional images (e.g. photos from a camera) is now commonplace. Smartphones come bundled with AI software that can process photos on the fly to automatically correct for lighting or tag people and landmarks. Industry uses AI Vision to detect defects and predict future failures without human intervention.

At the same time, the availability of hardware capable of acquiring 3-dimensional (3D) images is also commonplace. For example, the latest iPad and iPhones have integrated LiDARs enabling the capture of 3D images.

The current crop of image processing AI tools have been trained exclusively on 2D colour images and so they underperform when processing 3D images such as those generated by a LiDAR or X-ray CT scanner. The project will aim to combine transfer learning (network pre-training) with network architectures that can natively process image data in 3D. Thus, significantly accelerating processing times for Big Data (single file bigger than 1 Terabyte).

The successful candidate will have the opportunity to collaborate with scientist and engineers across Rutherford Appleton Laboratory and get trained on new software techniques and beamline operation.