Investigation and Benchmarking of FPGA AI Cores for Lens-Free Imaging Algorithms

Objectives

The objective of thisthesis is to investigate and benchmark the use of the Xilinx Versal AI Coreplatform for accelerating lens-free imaging algorithms.

The student will:

• Analyze the computational requirements of representative reconstruction algorithms for lens-free imaging.
• Investigate the Versal AI Core architecture, including:
• AI Engines
• Programmable Logic (FPGA fabric)
• Embedded processing subsystem
• Memory architecture and data movement
• Implement selected lens-free imaging algorithms on the Versal platform.
• Evaluate different acceleration approaches and partitioning strategies.
• Benchmark performance against CPU and/or GPU reference implementations.
• Identify architectural bottlenecks and optimization opportunities.

Research Questions

The thesis aims toanswer questions such as:

1. will lens-free imaging algorithms benefit most from Versal AI Core acceleration?
2. What performance gains can be achieved compared to conventional CPU or GPU execution?
3. How does power efficiency compare across implementation options?
4. What are the limitations and challenges when deploying computational imaging pipelines on heterogeneous adaptive SoCs?

Methodology

The student will:

Phase 1: Literature Review

• Review lens-free imaging concepts and applications.
• Investigate the Versal AI Core architecture and development tools.

Phase 2: Algorithm Analysis

• Characterize computational complexity, memory bandwidth, and parallelization opportunities for the selected algorithms.

Phase 3: Hardware Acceleration

• Implement selected kernels on the Versal platform.
• Explore multiple architecture mappings:
• CPU-only
• FPGA acceleration
• AI Engine acceleration
• Hybrid implementations

Phase 4: Benchmarking

• Measure:
• Execution time
• Throughput
• Latency
• Resource utilization
• Power consumption
• Energy per reconstructed image

Phase 5: Evaluation and Recommendations

• Compare implementation approaches.
• Identify best practices for future computational imaging systems.
• Provide design guidelines for deploying lens-free imaging on Versal-based platforms.

Expected Results

The thesis is expectedto deliver:

• A comprehensive performance analysis of lens-free imaging workloads on Xilinx Versal AI Core.
• Optimized hardware-accelerated implementations of selected reconstruction algorithms.
• Quantitative benchmarking results against conventional computing platforms.
• Recommendations for future imaging system architectures leveraging adaptive computing technologies.

Required Skills

Preferred background:

• Digital design and FPGA development
• C/C++ programming
• Embedded systems
• Signal processing or image processing
• Basic knowledge of AI/ML acceleration is a plus

Type of internship: Master internship

Duration: 6 months

Required educational background: Computer Science, Other, Electrotechnics/Electrical Engineering

Supervising scientist(s): For further information or for application, please contact Abelardo Gonzalez Diaz (< e-mail verwijderd om veiligheidsredenen >)

The reference code for this position is 2026-INT-317. Mention this reference code in your application.

Imec allowance will be provided for students studying at a non-Belgian university.

Applications should include the following information:

• resume
• motivation
• current study

Incomplete applications will not be considered.