Jump to content

CineFO

Members
  • Posts

    1
  • Joined

  • Last visited

CineFO's Achievements

  1. A few weeks ago I stumbled on a Youtube video from user Unnamed for a Reason (https://youtu.be/35VcYn3AobY) where the X-H1's eye tracking performance with the latest firmware update is demonstrated. I wasn't aware of the v2.11 update and it brought my attention to it. Chatting with him in the comments section he told me X-H1 users in a forum noted an improvement in eye tracking which prompted him to do the test, but he had already update the firmware and couldn't really provide a solid comparison. I loved his methodology for using movies as a tracking subject because it controls more variables, and since I hadn't updated the firmware myself, set out to perform the test in a similar manner. I had intended to record the results in Video Mode with AF-C and Single Shot with AF-S to compare face and eye tracking in the respective modes but with the hassle of setting it up, some trials and failed attempts over a few hours, I somehow neglected to record any Single Shot footage before the firmware update. Still, I reasoned the face tracking in Video Mode might reflect the kind of performance difference in Single Shot before and after the update. The default tracking settings were used so there may be ways to tweak them and get better results. Movie sequences were chosen for the wide variety of shots and some fast cutting, and also to see how close you can get to an actor before the face is no longer recognized (The Good, the Bad and the Ugly, and Once Upon a Time in the West), to test how it does with long takes (Kill Bill vol.1, Goodfellas), and seeing how the tracking fares with the face being obstructed every now and then (Pulp Fiction). The completed video comparison is a little over 25 minutes with some surprising results (the new firmware doesn't always outperform the old). The X-H1's LCD screen was recorded with another camera and the plane of focus is off sometimes. The settings are displayed for those who are curious. Some sequences were only recorded once using the parameters as the whole thing was time consuming enough as it is, and although the methodology is not 100% foolproof and scientific I thought the community here might appreciate what was captured. If there are any questions, feel free to ask in the Comments section. I make no claim over the recorded materials and am not trying to profit from the video but I do have some concerns Youtube may take it down due to copyright infringement. Without further ado, here ya go: https://youtu.be/WveCA-DLDA0 Cheers! -Paul
×
×
  • Create New...