Recent Projects
Covid-19 Predictions
We did some data analysis on the reported cases of Covid19, to try to make predictions about future cases of Coronavirus, as well as determining when it is going to peak and at what value the curve will flatten.
SubT 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.
PyTorch Object Detection and Tracking
Learn how to use Python and PyTorch to detect multiple objects in an image, and then how to track objects across video frames, using the Yolo V3 model. Full code and examples are included.
PyTorch Image Classification
An overview of training a deep neural network in PyTorch with your own pictures, and then using it for image classification and to perform basic inference on single images.
NLP Text Classification
A walkthrough of text classification in Python, using pipelines, NLP (Natural Language Processing), the NLTK toolkit, Tf-Idf vectorizing, XGBoost and other language techniques.
How far are we from real AI?
CYNET founder Chris Fotache and IBM Watson executive Karen Lomas talked to TechNative.IO about the future of AI, and how likely we are to have a real artificial intelligence any time soon.
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.
Building CYNET One
An overview of the hardware part of building our first drone. From choosing the frame and motors, to the Pixhawk autopilot and then adding a Raspberry Pi to run the controller software.
SLAM on a budget
This is how we were able to run SLAM (simultaneous localization an mapping) for a drone or robot, using just a cheap $150 Lidar unit, and Python libraries. Great results with very good speed!
There is no try
It's do, or do not. And for us, there's nothing we can't do.
Everytime we hear something is impossible, we take it as a challenge.
Projects in Progress
Smart surveillance camera
Using just a basic camera, we're trying to get information on what's happening in front of the house. A simple convolutional neural network will tell us what's out there (look, a car! a child! someone on the porch!), but adding k-NN over dynamic time warping will tell us what's happening (like "wife came home", or "children are playing in the frontyard", or "someone's walking up to the door").
Stock market predictions
A very simple task - look at past stock performance, and try to predict where it will go tomorrow. But by adding deep learning features, like LSTM and RNN, we believe we can get much more accurate forecasts. That, plus the feature selection, which include not just the particular stock, but many other financial indicators.
Get updates