Fujifilm Manager Interview (D.Etchells – PP): A Look into Fuji’s Deep Learning AF, Why X-T5 has No Battery Grip, Firmware Improvements and More
Manager Interview
David Etchells, formerly at Imaging Resource, and always highly respected for his deep technical knowledge, has now shared his talk with Fujifilm managers at the CP+ earlier this year.
As usual, David knows how to dig deep into technical details and make the right questions.
Here on FujiRumors we will provide just a short summary in bullet points for those of you who are in a hurry and want to give it a quick glance. But for the full talk you have to check out the very interesting article at petapixel.
Deep-learning based AF tracking system
- it’s an enormous task to develop it. Tens of thousands of photos of an intended subject must be manually processed by humans
- Fujifilm has to constantly add new elements to the training database to improve AF tracking
- X-H2S firmware 3.0 improved mainly
– AF tracking with fast moving subjects
– increased the AF frequency during continuous shooting by making the algorithm more efficient and processing faster (double the AF processing speed of previous firmware) - they talk how certain AF improvements might come to X-T5 and X-H2 at some point – which then happened on May 24
- Fujifilm might be the first manufacturer, who, with their latest firmware, now tightly couples AI subject recognition with conventional phase-detect-based AF tracking. David says that this is significant because AI subject recognition and conventional AF tracking are two fundamentally different processes. Getting the two to work together in close coordination is a big breakthrough.
Fujifilm X-T5 – 3 millimeters too much ;)