If you can imagine a Computer Vision system, we can make it.

In fact, we can make things you probably never imagined.

Here’s a video-based heart-rate monitor we made. Who’d have thought that was possible?

So, let’s see some of the types of Computer Vision there are to choose from, and then let’s look at what the prices are likely to be for your project.

When you’re ready, open a Live Chat with us using the bubble in the bottom right, and let’s discuss what you’d like to make.

The simplest CV task – find objects in images and videos. Specify what you want to see, and we’ll make you a model. Simple.
This is also the basis of most other model types, that identify a thing, and also tell you something about what it’s doing.

As long as there’s something of known length in the frame, we can tell you what speed things are moving.  Great for traffic, sports, wildlife, and many other things.
The speed can be in any units, and appears next to the object that’s moving. Can also be combined with counting.

This gives an estimated skeleton for all the people visible in the frame – and it works for animals too. Great for sports and animal welfare programmes, but also has many other uses, like art projects in which dancers control the tempo of music, rather than the other way around.

What if you want to know how many people walked past your store today, but you don’t want someone walking back and forth all day to be counted multiple times? You need our Counting and Tracking algorithm. Cars, animals, robots – tell us what you’re looking for and we’ll make a system to tell you how many there were, and where they all went. The numbers are next to the objects.

Since this is custom work, it depends. But you can figure out a ballpark. Here are the bits that go together to make the final price.

The  Model

The actual design and training of the model. This is the bit we specialize in, and it’s where you get the most value for the least cost. The more object types you’re looking for, and the more processes it’s going through after that, the more it costs, but we believe we give the best bang-per-buck in the industry.
The model will be fully documented. It’s a folder of Python and Weights files, by default, but we can also add Integrations, Interfaces  and Apps to it – see below.

Finding and labelling training data

Before we make the model, we need labelled training data. We can do that for you, or you can save money by doing it yourself.
There’s a free tool for that that we can give you. Generally 50 labelled images per class is about the right number, but we can discuss what’s right for you.

Integrations, Interfaces and Apps

We can connect your model to your other software, or even build entire Interfaces or Apps around it.
Depending on your existing software, this could be very simple or very complicated. Send us a Live Chat and we’ll help you figure that out.

Inferencing hardware

You can use your own hardware, which we can consult on, you can hire cloud hardware, or we can manage the whole thing for you.
We use vast.ai for inferencing, which is very cost-effective for heavy workloads. For lighter workloads, we tend to use AWS. Open a Live Chat and we can help you figure that out.

Discounts

We’re building a platform that’s intended to eventually have hundreds of computer vision models available off the shelf. If you’re happy to let us sell your model on once you’re finished with it, we can give you a discount.

So, to put some firm details on that, open a Live Chat using the box in the bottom right and we’ll talk about it.