The Fastvideo Formula: How to be an Innovative Company in GPU-based Image Processing

Serzhenko Fyodor, CEO at Fastvideo tells us about GPU image processing.

First of all, how are you and your family doing in these COVID-19 times?

Serzhenko Fyodor: I suppose all of us are having exceedingly difficult times, even those of us who were lucky enough not to get sick. Our families, friends, relatives, and coworkers are still in danger and we have to follow public health regulations and be patient. Both the pandemic and the (first, second, possibly third?) lockdowns have already changed our lives, and we have to be flexible enough to be able to cope with the arising challenges.

Tell us about you, your career, how you founded Fastvideo.

Serzhenko Fyodor: Before starting my entrepreneurial career, I had a strong background in applied physics and math, as well as hands-on experience in software and business development.

Fastvideo was founded in 2005. Since then, we’ve invented lots of interesting things. We used to manufacture high-speed cameras and software for industrial and scientific applications. Starting in 2010, we switched to GPU-based image processing software based on NVIDIA CUDA technology. We believed that the future of parallel computing would be promising. It looks like it was the right move.

How does Fastvideo innovate?

Serzhenko Fyodor: We’re seeing a growing demand for high-performance image processing solutions for industrial and scientific applications. For example, we pioneered the first parallel JPEG codec on GPU with a breakthrough performance which was not possible either on CPU or FPGA. Since then, it’s been the fastest existing solution for JPEG encoding. What’s more, we implemented an image processing library for NVIDIA GPUs to perform the full image processing pipeline and get high quality, excellent performance, and low latency. It works on all NVIDIA GPUs: mobile (Jetson), laptop/desktop (GeForce), and server (Quadro/Tesla).

Our solutions are implemented by lots of companies with well-known names. Among our customers are startups as well as F500 companies. In recent years we participated in projects like Nokia’s OZO, plus created software for light field cameras from Lytro, Yahoo’s Flickr service, the news feed at VK.com, and others. 

For example, our software processes 400,000 images per second on VK.com servers. This kind of performance was achieved on NVIDIA GPUs with our highly optimized software. You can find even more case studies and application examples of our software on our Fastvideo blog. 

We are proud to cooperate with customers from various industries. The majority of our customers develop complex end-user solutions involving video cameras as a part of the solution. They require high-performance image processing, which can be beyond their customers’ experience.     

For example, one of our customers processed RAW images in Adobe CameraRaw software. We suggested a GPU-based solution which turned out to be more than 30 times faster.

Though we are focused on GPU programming, we do not develop AI/DL applications because there are a number of solutions based on conventional image processing algorithms with better performance characteristics. High performance is the key for these applications. As for Fastvideo, we provide software optimization to accelerate image processing tasks. Our solutions are designed to preprocess raw images for further neural network training.

How does the coronavirus pandemic affect your business finances?

Serzhenko Fyodor: The COVID-19 situation has affected some of our projects: they were postponed because the customer’s projects involved hardware (video cameras and servers), so it was not possible to resolve issues remotely.

However, we’re sure it’s time to innovate, and most of our customers agree with us; they prefer not to put their projects on hold so they’re working from home, because actually there are lots of software development tasks which can be performed while we stay in the comfort of our apartments and houses. 

Though the negative trends are continuing, we can clearly see a trend towards remote work and automation, and that is what we’ve been working on for many years.

In response to the current situation, last year we built a software solution for remote color grading and added it as a part of our product portfolio. It’s intended for customers working remotely in the VFX and digital cinema industries. The software is designed to provide a high-quality color reproduction service via network: it comprises the Fastvideo high-performance J2K codec based on NVIDIA GPU and a reliable data delivery solution. The key feature of the solution is our highly optimized core engine for GPU image processing. Now our customers can perform color grading tasks remotely, so they don’t need to be in the office.

Did you have to make difficult choices regarding human resources and what are the lessons learned?

Serzhenko Fyodor: Our employees are the key value of our business, so we strived to preserve our core team, though we had to terminate some projects which we considered to be irrelevant to the current situation.

We reexamined our approach to business development and customer relations to increase business efficiency and find our own path in the current situation. We sincerely hope that for now, we’ve done our best, but we are ready to face new challenges.

Our company has always been committed to remote work, and from our own experience, we know that anything is possible if you trust your team members. To overcome the lack of contact we held regular team calls and supported each other as much as we could.

How did your customer relationship management evolve? Do you use any specific tools to be efficient?

Serzhenko Fyodor: As we are a B2B company, our main sources of new customers are exhibitions, in-person meetings, organic traffic from search engines, and LinkedIn connections. Unfortunately, offline exhibitions are not being held currently, and we don’t believe that virtual exhibitions have the same impact on customer relations. Although we can’t meet with our customers in person, we’ve been successfully contacting them remotely, and the results of those meetings are more than satisfying. 

We believe that a software company like ours should educate its customers and demonstrate its expertise, so we put a lot of effort into a content generation for our Fastvideo blog. It helps us attract new subscribers to our LinkedIn group, engage our existing LinkedIn connections, and convert them into paying customers. 

One more source of growth for us is well-timed software improvements and further development of our solutions so we are ready to meet new customer requirements. 

Did you benefit from any government grants, and did that help keep your business afloat?

Serzhenko Fyodor: Recently we received funding from the Russian government to develop GPU-accelerated software for computed tomography. It’s a significant help in the current situation, for which we are incredibly grateful.

Our core engine software can be utilized in various applications, and we’re currently in negotiations with companies from industries we have never worked in before. Taking this into consideration, the COVID-19 situation has forced us to put more effort into business development.

Your final thoughts?

Serzhenko Fyodor: The current situation in the world and Russia is more than just a pandemic: I believe that all of us face a severe economic crisis in combination with the pandemic. The cumulative influence of these two factors is immense, so the only advice I have for all of us is not to stop searching for new opportunities and to do our best to decrease the effect of the two negative factors.

I’m quite confident when I say that we are adapting to the norms of the new reality and expanding our activities to overcome the crisis. The year 2020 was challenging and unpredictable but taught us a lot, and we still have the strength and courage to move forward.

Your website?

Fastvideo website – https://www.fastcompression.com 

Fastvideo blog – https://www.fastcompression.com/blog/content.htm