VISIA aims to restore visibility in challenging weather conditions such as rain, fog,
and haze to ensure safe and reliable perception for autonomous and assisted-driving systems.
By removing environmental degradations while preserving scene structure and colour fidelity,
VISIA enables vehicles to better understand their surroundings even in low-visibility scenarios.
A core objective of VISIA is real-time deployment on resource-constrained edge devices.
The project focuses on developing lightweight and efficient restoration models that can process dashcam footage live on the vehicle,
ensuring uninterrupted perception and enhanced safety without relying on cloud computation.
One of the key capture resources is IIITH’s own Bodhyaan platform. This vehicle has been widely used video capture for multiple projects and is used for Project VISIA. In addition, we are also developing a Mobile phone-based App as a low-cost option for capturing data and uploading it to our cloud.
VISIA enhances dashcam footage under rain, fog, and haze conditions using lightweight deep learning models deployed on edge devices like NVIDIA Jetson. The system delivers real-time visibility restoration, enabling safer perception for autonomous and assisted driving.