Teach on your own the Eocortex neural network to solve unusual tasks
EVERY
PROJECT
IS UNIQUE
That is why we are planning to create an online service that will allow our partners to independently train a neural network to identify any object types.
We are going to provide a tool for solving unique cases
Construction
Agriculture
Urban infrastructure
Traffic
Livestock
Manufacturing
How it works
First step
01
Collecting materials
Collect photos and videos of the required objects at the site. It is important to consider the real system operation conditions: weather, time of day, camera angle, camera position, etc.
Upload photos and videos to our service. Highlight the desired objects (mark them up) in the first 50 photos. Service will automatically check the markup and point out any errors. When the check is done, mark up the remaining images.
Within two days the Eocortex online service will train the neural network to find the required objects in the frame. Before any payments you will have an opportunity to check the accuracy of the recognition directly on the service. In the result you will get the neural network file and one license for it.
Takes training of the neural network by the service
~3 days
Up to 2 days
TIMELINE
You can get a working prototype within 5 to 7 days
Take full advantage of the service
Exclusiveness and availability
Only you will have access to the trained neural network. Once trained, you can use the same neural network in other projects.
Free-of-charge updates
You are not bound to the specific Eocortex version. All updates starting from 4.3 will be available and free of charge
Simple training process
No need to have an in-depth knowledge of machine learning. We'll give you a tool that makes it nice and easy.
Fast testing
We'll train a neural network to distinguish your type of object in 2 days. The accuracy depends on the task, set of samples, and quality of the markup.
WHAT
YOU'LL HAVE
TO PAY FOR
ACCESS
Paid only once. After payment you will get permanent access to the service
TRAINING
The first training of a neural network is free of charge in the pilot version. Each subsequent one will require individual payments.
NEW PROJECTS
You can use a trained neural network in other similar projects. Only use in new systems will be charged separately.
Terms for the first service users
Participate in product testing and train a neural network on favourable conditions
Cost for the first 6 partners (testing and long-distance use)
Cost after the launch of the online service
ACCESS TO THE SERVICE (unique login and password)
$ 0
$ 2 500
FIRST NEURAL NETWORK TRAINING (1 neural network and 1 license)
$ 0
$ 800
NEURAL NETWORK RE-TRAINING (if accuracy improvement is needed)
$ 500
$ 600
USAGE IN A NEW PROJECT (if a previously trained neural network must be used in a new project)
On demand
On demand
* Licensing cost and conditions may change as the product is still under development.
FAQ
How long will it take to mark up objects in all photos or video footage ?
Depends on the number of images. Usually it takes 2 to 3 days to mark up 2,000 photos
Can I use an already trained neural network in another similar project ?
Yes, you can! Previously trained neural network can be used in several projects at the same time. But for each new project you will need to buy licenses to use the trained neural network.
Can Eocortex use a trained neural network for its own commercial purposes?
No. We are not allowed to use your markup and materials for commercial purposes.
What if the neural network needs to be trained again, in case the accuracy is not good enough ?
You will have to pay a separate fee for repeat training (re-training)
Is a system with a trained neural network a custom (dedicated) version ?
No. All new software updates will be available and free.
What if the customer has provided only video footage from the site?
Simply upload your video to the service. It will automatically separate the video into frames at the required frequency
1. Upload photos or videos with the desired object to the service. Preferably, there should be several objects in the frame
2. If a video is uploaded instead of a photo, our service will automatically split it into frames at a proper frequency
3. All that remains is to select the desired object in each photo with the mouse
* We plan to add new tools in the future to reduce the markup time
The most important
Photos and videos must be collected under real system operation conditions (Images from the Internet would not fit)
* You can use the same trained neural network in other projects. Just purchase an additional license for that without the need to re-train the neural network.