The problem HazeXpert solves

HazeXpert was mainly designed to help firefighters. Since their vision is what saves people's lives, a little boost to their vision can effectively save more lives. HazeXpert dehazes/desmokes video in REAL-TIME !! This is the first of its kind and achieves latency of 0.3 seconds on a laptop. With specialized GPU, this can be made better.

Challenges we ran into

Initially we tried to train a Machine Learning model but It had immense latency and defeated the purpose. So we had to start all over again and this time we utilized the OpenCV module for all out image processing needs. Later we realized that although the latency significantly reduced, it was not what we were hoping for. After a few sleepless nights, we found a solution that brought down the latency to 0.3 seconds. Instead of processing the video frames and then sending the video back, we decided to stream it and use pipelines and various other features for faster processing.

Ultimately, HazeXpert was made Real time, with about 0.3 s latency when started and an average of 0.5 seconds of latency, on a laptop without a GPU . Imagine how fast it could be if we had access to laptops with GPUs