Well this week was a little bit of a wild card and our group had to do some serious digging/rethinking about our project as a whole and our project idea’s. So I started off this week by trying to find some kind of metric to measure the sharpness of an image which then lead me down a rabbit hole of reading some masters papers on video deblurring and some of the techniques available. This then lead me to a paper talking about doing exactly that; video deblurring (paper can be found here). In this paper they talked about how the standard time for a single frame to be deblurred is usually in the minutes, closer to 1-2 minutes, to which they proposed a ML model that can do it an about a second. This essentially meant that our original idea of doing this video processing in real time would have been nearly impossible, so we had to decide if we wanted to pivot or continue on with the project.
We decided that it would be in our best interest to go forward with the project, but this time do the video processing after the video has already been recorded. This way we can use the technique this paper is talking about without having to worry about making any massive design changes to increase the speed. We are moving forward with the research stage of our project according to our plan. We have some great leads for getting a bare bones model up and running so that we can start to make changes to the model and play around with its efficiency. Hopefully next week we can do some more research and talk with some faculty about our approach.
0 Comments