Implementation of Video-Dehazing Based on AOD-Net Using Docker on Jetson Nano

This paper deals with the implementation of a machine learning model used for dehazing video inside a Docker container. Docker containers provide the user a standardized environment, promoting portability and making it easier to deploy the model across various computing platforms. The paper discusses in detail about the decision behind selecting the model, process of containerizing the dehazing model, including required dependencies and configurations. The advantages of usage of Docker for video dehazing are discussed, stressing upon the enhanced portability, streamlined deployment, and potential for cloud-based applications. The performance of the model is judged based on parameters like PSNR, SSIM and average time taken for processing each frame. The paper concludes by outlining the effectiveness of Docker in facilitating the practical use of machine learning models for video dehazing tasks.