Post-master's student

Project
PREP0003757
Overview

The NIST Information Technology Lab (ITL) and Engineering Lab (EL) are collaborating on a project for real-time image processing for Additive Manufacturing. To handle real-time constraints, computations on Field Programmable Gate Array (FPGA) devices will need to be enabled, likely involving both traditional Computer Vision algorithms and Deep Learning models.

We plan on instrumenting a hard real-time system that can meet the time sensitive deadlines for detecting sparks from a high-speed camera that is monitoring the interaction between the melt pool and laser. There are three methodologies to consider. 

1. The camera contains a built-in FPGA that can process images as they are captured. 

2. The capture card has a slightly higher-end FPGA. 

3. The capture card can transfer image data into system memory, allowing the host system to process images using either the CPU, GPU, or a combination of both.

To this end, we are seeking a Computer Scientist who will focus on developing algorithms to process frames in real-time from a high frame rate camera. The processing algorithms may utilize the camera’s built-in Field Programmable Gate Arrays (FPGA), the capture card’s built-in FPGA, or traditional computer CPUs and GPUs.

Computer Vision AI models for Additive Manufacturing image processing

Qualifications
  • A completed or in-process graduate degree in Computer Science, Engineering, Manufacturing, or a related field.
  • 1—2 years of relevant experience.
  • Familiarity with image analysis algorithms.
  • Familiarity with FPGA programming.
  • Familiarity with CPU and/or GPU image analysis.
  • Experience with AI/Deep learning workflows, such as LSTMs.
Research Proposal

Key responsibilities will include but are not limited to:

  • Develop image analysis algorithms that target the highspeed camera’s FPGA.
  • Develop image analysis algorithms that target the capture card’s FPGA.
  • Develop image analysis algorithms that target the traditional computer’s CPU(s) and GPU(s).
  • Measure real-time throughput for developed image analysis workflows.
  • Create AI/Deep learning workflows for training AI models for analyzing images in a series.
NIST Sponsor
Derek Juba
Group
Information Systems Group
Schedule of Appointment
Part time
Start Date
Sponsor email
Work Location
UMD Campus
Salary / Hourly rate {Max}
$38.35
Total Hours per week
20
End Date