Our team can implement AI in real time on FPGA’s and SoC’s by partitioning software and hardware to achieve real time inference. In this way, we can perform advanced analysis on the edge with very low latency without requiring cloud infrastructure nor communications to the outside. A typical system comprises:
- Serial interfaces and field buses to read data from sensors and PLCs.
- AI on the edge:
- Time series analysis and deep learning for predictive maintenance
- Computer vision and deep learning for automated quality inspection
- Cloud connectivity to upload metrics
- Dashboards to display insightful information
We use real time deep learning inference on the edge for multiple applications:
- Predictive Maintenance: we develop systems that read data from sensors that measure the performance of industrial assets, process this data using deep learning techniques (LSTM, autoencoders) to detect anomalies and predict failures before they occur.
- Quality inspection based on computer vision: we develop systems that processes images using Convolutional Neural Networks (CNN) to detect defects and classify objects according to their physical features (size, shape, color)