Vaaman

The reconfigurable edge computer

Available for pre-order

View Purchasing Options
Feb 27, 2025

Project update 2 of 5

Combining Our FPGA Acceleration With the Hailo AI Kit

by Akshar Vastarpara

We are excited to share this groundbreaking update regarding our Vaaman platform, which can now be enhanced with the Hailo AI Kit. This integration exemplifies how FPGA technology remains highly relevant alongside AI accelerators. By combining AI capabilities with FPGA-based computation, we unlock new possibilities for:

Pushing the Boundaries: Performance Benchmarking

As part of our ongoing commitment to transparency and performance optimization, we conducted extensive benchmarking on Vaaman using Hailo8L (13 TOPS). Below are the inference results across various AI models:

ModelFrames Per Second (FPS)Latency (ms)
RESNET5023.9930.17
MNV11866.581.85
MNV21738.53.4
MNV374.7211.69
MNV357.29NS *
MNV46.06NS *

* - NS: Not supported due to NMS network limitations

These benchmarks highlight how AI models run efficiently on Vaaman with Hailo8L, demonstrating its suitability for high-speed inferencing applications across multiple industries, including robotics, automation, and edge computing.

Why FPGA is Still a Game-Changer?

Despite the rise of dedicated AI accelerators like Hailo, FPGA technology continues to play a crucial role in computational applications. Here’s why integrating FPGA with AI accelerators makes sense:

1. High-performance DSP Processing

Digital signal processing (DSP) is fundamental in industries such as telecommunications, audio processing, and radar systems. FPGA can be optimized to handle DSP workloads with low latency and high efficiency, making it ideal for applications that demand real-time performance.

Example: In a smart hearing aid, FPGA can filter out background noise in real time, enhancing the clarity of the user’s experience.

2. Fast Fourier Transform (FFT) Acceleration

FFT is essential in various signal processing applications, including medical imaging, communications, and industrial monitoring. FPGA-based implementations can significantly accelerate FFT computations, enabling faster data analysis and improved response times.

Example: In MRI machines, FPGA-accelerated FFT can help reconstruct images faster, reducing patient scan times and improving efficiency.

3. Low-latency Control for Real-time Systems

Autonomous driving, robotics, and industrial automation require precise, real-time control. FPGAs provide deterministic execution, ensuring millisecond-level response times, which is critical for applications such as:

4. Custom Hardware Logic Implementation

FPGAs allow users to create custom processing pipelines tailored to their specific computational needs. Unlike fixed-function processors, FPGA logic can be reconfigured on-the-fly, making them adaptable to various workloads without the need for additional hardware investment.

Example: A high-frequency trading system can use FPGA logic to execute trades in microseconds, reducing latency and maximizing profits.

Vaaman: Bringing AI and FPGA Together

Vaaman offers a seamless fusion of AI acceleration with Hailo and FPGA-based computing power, providing an unmatched level of flexibility and efficiency. Whether you are working in:

Vaaman ensures that your applications are optimized for peak performance.

This unique combination makes Vaaman a versatile choice for engineers, researchers, and developers looking to push the boundaries of innovation.

What’s Next?

We are continuously working on expanding the capabilities of Vaaman. Future updates will focus on:

Stay tuned for further updates, and thank you for supporting our journey towards redefining the future of computing! 🚀


Sign up to receive future updates for Vaaman.

Subscribe to the Crowd Supply newsletter, highlighting the latest creators and projects