This week I continued research on IO for the Ultra96. I was able to find example code for video in and video out, which I will need to modify to work with video file in — the example used video stream in to 1080p@60FPS video out. In looking at the specs of the board, and the available data for training that Joshua found, we decided to change our spec framerate to 30FPS. There are more video datasets at 30FPS than our original idea of 24FPS; this was because of the difference between “p” and “i” formats which we initially overlooked. The “p” formats are more common on pixel screens and “i” format is from when video was still interlaced.
I was also able to get communications setup between the ARM core and FPGA of the Ultra96. This ended up being a prerequisite to some setup I had to do for looking at IO, so in this case, I was able to check off something that we planned to be further down the line in our gantt chart, which is always a good thing to be able to do.
I started our slideshow for the design presentation and began the draft for the report as well. I have been working on those, and running metrics to have hard data to present on those as well, specifically runs comparing MSE to SSIM to VMAF to motivate the need for VMAF as a metric over the other two more well-known ones.