Autofocus

The E-M1X is billed as a sports, action and wildlife camera and has the burst speed to prove it. But what good's burst speed if your autofocus system can't keep up? Fortunately the E-M1X does a decent job keeping up with moving subjects, even at 18 fps, and has impressively sticky subject tracking.

The camera also offers machine learning-driven AF modes for planes, trains and automobiles. We didn't get a chance to test it on trains, but it proved very reliable at identifying and tracking planes and motorcycles.

Takeaways:

  • For best results in AF-C, max out the 'C-AF Sensitivity' to +2
  • Doing so increased the single-point hit rate of our bike test from 65% to 80% and decreased the degree to which images were out-of-focus
  • The E-M1X falls behind the competition in terms of overall focus speed and accuracy
  • AF tracking is impressively sticky
  • We had a hit rate of close to 90% using AF-C+Tracking, even with AF-C sensitivity left at default
  • Machine learning-driven AF modes are neat, but limited to only a few use cases

C-AF Sensitivity settings

The E-M1X has five 'C-AF Sensitivity' settings ranging from +2 to -2. Olympus claims users should increase the sensitivity for subjects moving quickly toward or away from the camera and decrease it for relatively static subjects. The later can be used to avoid the camera re-focusing if your intended subject has other subjects passing in front. But for the most part we found increasing the sensitivity to +2 increased our hit rate not just in the bike test, but in most real-world shooting scenarios.

AF-C Performance

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Default C-AF Sensitivity - shot at 10 fps (mechanical shutter) using the 40-150mm F2.8 lens.

This first test takes a look at how fast the camera can refocus on an approaching subject using a single point. Using the camera's mechanical shutter (10 fps) and default AF settings (C-AF Sensitivity set to 0) the E-M1X's hit rate was roughly 65%. About one third of the time focus fell behind the subject, only to catch back up later. We also saw similar results when using the faster e-shutter (18 fps): AF keeps up about two thirds of the time and falls behind the other third of the time, eventually catching back up.

These aren't particularly impressive results for a product geared toward fast action, especially considering the camera's competitors. Fortunately you can bump that hit rate up closer to 80% by increasing the C-AF Sensitivity to its maximum, +2. Not only does increase the number of in-focus shots, the 20% of images that are out of focus are noticeably less mis-focused than when the sensitivity is set to 0. However, even at +2, we're still seeing some back-focused shots, which is concerning, considering the subject was not moving very fast.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Maximum C-AF Sensitivity - shot at 10 fps (mechanical shutter) using the 40-150mm F2.8 lens.

Tracking performance

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Default C-AF Sensitivity - Shot at 10 fps (mechanical shutter) using the 40-150mm F2.8 lens.

This second test takes a look at how well the camera can identify and track a subject moving around the frame, on top of the challenge of assessing distance and driving focus to the correct point. This is a more complex challenge, so usually gives a lower hit-rate than the simple AF-C test. Very unusually, that's not what we saw here.

Using the mechanical shutter (10 fps) the E-M1 X's hit rate for our weaving bike test was 90% or better. Never did the camera entirely lose our subject and the frames not in focus are only off slightly. This is impressive.

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16

Default C-AF Sensitivity - Shot at 18 fps (e-shutter) using the 40-150mm F2.8 lens

Our hit rate with the e-shutter (18 fps) is about the same, around 90%. Again, the camera’s only falling behind the subject from time to time and rarely losing him completely. It's worth noting, both weaving examples were shot with C-AF sensitivity at its default of 0.

Machine learning-driven AF

Shot using airplane tracking mode.
ISO 1000 | 1/3200 sec | F7/1 | Shot using the Olympus 300mm F4 Pro

The E-M1X is among the first cameras on the market with AF modes trained to identify specific subjects. Machine learning is increasingly being used by manufacturers for more accurate face and eye detection, or to identify animals and other objects, in real-time. At launch the E-M1X can detect trains, planes and automobiles.

Users must first set which subject to look for from within the camera's AF menu (you can only set one at a time). Then, when said subject shows up in the frame a green box will appear around said subject, confirming it has been identified. For both motorcycles, cars and planes, we found the camera identified and locked onto subjects with a good deal of accuracy.

These modes are smart enough to not only track the outline of say, someone riding a motorcycle, but actually focus on the rider's helmet, or the cockpit of the plane. For scenes with multiple planes or motorcycles, the camera will settle for the largest in the frame. However, if you'd like the option to toggle between subjects, you can leave the camera in Single Point and manually place your point over the further subject - the camera will then prioritize tracking them instead.

Shot using automotive tracking mode.
ISO 1250 | 1/1000 sec | F4 | Shot using the Olympus 12-100 F4 Pro at 50mm equiv.
Photo by Carey Rose

While these modes open up a ton of compositional freedom by allowing you to concentrate on framing rather than focusing, they aren't perfect: even on a clear day photographing planes, we got some soft shots. And the camera had a tendency to give up on tracking when foreground elements momentarily blocked the subject, forcing you to re-engage when the subject re-appears.