Face detection using FPGA
This represents our pipeline to create a object detection model like Yolov3 with its implementation on edge device which in our case is FPGA. Throughout training, optimization and implementation of given model we deliver you a final model that runs on target device and performs Face Detection alongside results at which FPGA performs.
Skin Tone Estimation
This system for skin tone estimation is a useful tool for beauty products recommendation. In addition to the estimation model itself, the system contains image preprocessing tools which makes it robust in all lighting environments.
Causal inference for root cause analysis in biotech
On top of its anomaly detection abilities AnomalAIzer offers state or art causal inference engine with ability to analyze complex multi-variate data streams enabling real-time tracking and response to data changes.
Improving deep learning model performance with synthetic data
Machine learning model performance is tightly connected to the quality of the data used in the training process. By using syntehtic data, the quality and the size od the training data can be greatly expanded.
Anomaly detection in multivariate data streams
AnomalAIzer is an advanced machine learning platform that allows to monitor and analyze complex multi-variate data streams. It offers real-time anomaly detection and visualization providing insight to workers in various scientific and industrial fields.
Vision-based navigation in GPS-denied environments
This localization and navigation pipeline allows UAVs to rely on their sensors to stay on track when the GPS signal is unreliable or unavailable. It is based on Robot Operating System (ROS) which makes it deployable on many platforms and can be connected to standard flight stacks.
On-board anomaly detection on OPS-SAT satellite
Satellites need to constantly monitor their telemetry to make sure everthing is working as expected. Anomaly detection is a core functionaly needed for any mission to ensure the longevity of the satellite.
Enhancing RGB images with multispectral data
Our deep learning model brings multi spectral data to standard RGB images, providing valuable spectral information without the need for expensive specialized cameras.
Vision-based On-Orbit Inspection for Optimal Solar Panel Deployment
Multispectral cameras are expensive and hard to acquire, especially the ones with proven flight herritage. Creating a module that enables multispectral measurements would help bridge the gap between functionality and cost-awareness.
Classification of Multispectral Land Cover Images Using FPGAs
Creating an AI model that can utilize multispectral images for classification tasks for Earth observation is relatively simple. Creating an AI model that can be run on the satellite whilst adhering to power and resourse utilization is a whole other story.