OpenVINO Review

A free toolkit providing improved neural network performance


What is our primary use case?

I created a retail recognition custom model — a model on RPX 2017. Afterward, I transferred it to OpenVINO for object detection and retail detection.

We have a team of three people who use OpenVINO.

How has it helped my organization?

It's a great solution for getting camera images, processing them, and extracting the reserves. It's better and more cost-effective than using Intel myriad X.

What is most valuable?

The inferencing and processing capabilities are quite beneficial for our requirements.

Compared to Jetson Nano or Jetson TX2, or Jetson Xavier, OpenVINO is a much more cost-effective solution. Processing-wise, they are comparable to Jetson and maybe Jetson Xavier NX.

What needs improvement?

The model optimization is a little bit slow — it could be improved. They should introduce some type of deep learning accelerator, like Jetson Xavier NX.

There is a lacking in vehicle recognition — types of vehicles. Differentiating between cars, SUVs and different types of light, heavy, and medium trucks can be tricky. We have to train such models ourselves and then transfer them onto OpenVINO.

For how long have I used the solution?

I have been using this solution for roughly 18 months.

What do I think about the stability of the solution?

It works quite well, but regarding the frames per second — If we want to add on a few more cameras to the hardware, we can't do that.

By assigning different IPs to our own software, we can access different cameras on a real-time basis, to analyze the traffic from different lanes of vehicles and roads.

This solution has never crashed and we haven't experienced any bugs; however, there are improvements that need to be made.

What do I think about the scalability of the solution?

I didn't scale it, but I want to build more than one board, on top of each other, so that we can improve the processing compute capabilities.

How are customer service and technical support?

The technical support is very good.

How was the initial setup?

I didn't experience any difficulties when setting up this solution.

What about the implementation team?

We have a really small team of about two to three people. One person has to go to the site and set up the hardware so we can access the board remotely. We can work from our office or any location. It's a one-time installation process; from there, our team can work from the office.

We were planning on deploying this solution on different devices, so the planning aspect took a little bit of time, but overall, we were up and running within a week.

What's my experience with pricing, setup cost, and licensing?

We didn't have to pay anything for Intel OpenVINO, everything was available on their site. All of their solutions, inference engines, and other model optimizations are all available for free. 

We didn't have to pay for any licensing with Intel OpenVINO. Everything is available on their site and easily downloadable for free.

Which other solutions did I evaluate?

We are currently trying Nvidia and Xavier NX because some of their models and capabilities look promising.

We are comparing the processing and usability of these two boards and we are planning to have a model GPU board with us also.

What other advice do I have?

I would absolutely recommend OpenVINO. I think it's a good introduction to machine learning and inferencing.

On a scale from one to ten, I would give OpenVINO a rating of nine.

Which deployment model are you using for this solution?

On-premises
**Disclosure: I am a real user, and this review is based on my own experience and opinions.
Learn what your peers think about OpenVINO. Get advice and tips from experienced pros sharing their opinions. Updated: September 2021.
535,919 professionals have used our research since 2012.
Add a Comment
ITCS user
Guest