The Challenge
Real-time video processing is computationally expensive. The challenge was to create a dehazing model that was both accurate and efficient enough to run on low-power edge devices like a Raspberry Pi.
Our Solution
To build an end-to-end system for real-time image dehazing, with model inference happening on a self-hosted API and video processed on an edge device.
Technologies Used
TensorFlowOpenCVRaspberry PiWebSocketPython
Key Results
Proved capability in building complex, end-to-end AI systems that span from cloud infrastructure to edge computing.