The AI Company #ThinkDeep

Artificial Intelligence research.
Integrating Human and Machine to augment reality.
Teaching machines to see the world the way we see it.

SubT Challenge Qualifier

Team was selected by DARPA (the Defense Advanced Research Projects Agency) as a qualifier in its ground-breaking SubTerranean Challenge (SubT). SubT is a multi-million dollar competition that will task teams of robots with autonomous exploration deep beneath the surface of the Earth.

After a few months of developing software for the qualification process Team was one of the only 8 teams across the world that continued through the virtual track of the competition. Read more details about our participation in SubT.

Team CYNET SubT Challenge DARPA

What We Do

Artificial Intelligence research
Artificial Intelligence

Using machine learning and especially reinforcement learning to develop agents that get smarter in time

Robotics robots and drones
& Drones

Researching the autonomous vehicles of the future, from flying robots to self-driving cars

Deep learning neural networks

Harvesting the power of deep neural networks so the computer can see the world the way we see it

Computer vision OpenCV

Object recognition, 3D scene modeling and everything that helps a robot perceive the environment

Teaching machines to
see the world the way we do

We're not describing each object to the computer. No more detecting lines, circles and textures. We just show our agent lots and lots of pictures and let it learn on its own. We do this using deep learning techniques, building neural networks with multiple convolutional layers, that can then be queried from portable or embedded systems.

Recent Projects

DARPA SubT underground robots
DARPA SubTerranean Challenge

Our team is competing in DARPA's SubTerranean Challenge, consisting of deploying a swarm of robots in an unknown underground environment, with the purpose of mapping it and detecting artifacts. We're using the ROS framework and Gazebo simulation.

Autonomous drone
Autonomous drone

A work in progress, our quadcopter uses a Pixhawk autopilot, with an onboard computer that guides it autonomously. The possibilities are endless. Use the front camera for obstacle avoidance, downward camera for navigation, GPS for creating flight paths, and more.

Land features detection in aerial video
Land features detection

We trained the Inception convolutional network model to classify thousands of aerial images. Then by running inference on the neural model, we can accurately detecting the land features below the drone. This helps by identifying landing spots or safe flight corridors.

On the quest
for the Artificial
General Intelligence

We want to go big and go bold! Like every AI researcher out there, we're looking for the Holy Grail of technology - the AGI. We believe that with exponentially growing computing power and the endless availability of big data, the path is set and in the coming decades one of us will get there.

Our Favorite Technologies

We're researching many exponential technologies, but here are just a few of the tools we've been using recently and what we're doing with them

  • PyTorch

    Our favorite deep learning framework, that we combine it with the power of Nvidia GPU's to train deep neural networks on large sets of images

  • OpenCV

    The standard in computer vision, we use it for image pre-processing, visual data augmentation, and basic shape detection tasks

  • Drone hardware

    We aim to build fully autonomous drones, with path planning and obstacle detection, by interfacing our computer with a Pixhawk autopilot

  • Microsoft Azure

    A lot of our compute hardware resides in the Azure cloud, as embedded IoT devices on drone and robots connect to it to execute more complex operations

  • Mobile apps

    As smartphone are ubiquitous now, we provide Android interfaces to most of our application so you get the power of Artificial Intelligence in your pocket

  • Python, C, Java

    ...and more. We choose the right programming language for each project, from Python from fast development to C, C# or C++ for faster processing

Get updates