Proving that Olympus hasn’t been sitting around twiddling its thumbs during the fairly long hiatus between new camera launches, the newly-launched OM-D flagship introduces some interesting new technologies and some interesting extensions of existing ones.
As a point of historical record it should be remembered that active sensor cleaning, in-body image stabilisation and multi-shot high resolution capture via pixel shifting were all pioneered by Olympus and the E-M1X introduces a couple more. At the top of the list are the three ‘Intelligent Subject Detection AF’ modes which employ ‘deep learning’ technology to adapt the autofocus tracking to specific type of subject matter. Deep learning is a form of artificial intelligence and the control algorithms have been derived from the data of tens of thousands of reference images of each subject so that the autofocus can react to any changes in direction, speed and distance from the camera (i.e. subject size). In the case of the E-M1X, it has to be able to do this at up to 18 fps which is why the camera has twin quad-core processors for data processing. The first three intelligent tracking modes are motorsport (i.e. cars and bikes), aircraft and trains. Obviously a lot of work goes into the development of these modes – Olympus says it’s been working on the concept for over three years – but it’s safe to assume there will be many more of them available down the track via firmware upgrades.
Does it work? You betcha. At a pre-launch event for Australian media and retailers in the NSW Hunter Valley, we were able to try out intelligent tracking firstly with a pair of rally cars and then a very fast (and small) aerobatic aircraft. The rally cars were running around a small gravel stage which included an old quarry bed so there were a number of high vantage points where they could be tracked for quite some distance, including through a sweeping curve. As the car emerged into the quarry, side-on to the camera, the tracking locked on instantaneously with a target rectangle that roughly covered its shape at that point. As a car started to turn into the bend, so presenting a smaller and different-shaped profile, the tracking target changed shape accordingly (and position, of course) so it stayed locked on to the subject… which is very impressive given the speeds involved. Then, as the car headed towards the camera position – always a challenge for autofocusing – the target again changed shape and size. Interestingly, it’s the subject-recognition algorithms that are doing all the work here and not the autofocusing system… or at least not the distance determining part of it. The AF system is essentially only focusing the lens using data from the intelligent tracking processing, rather than distance data from the sensor’s phase-difference detection pixels… which is why it’s so very much more reliable than anything we’ve seen before. It is really recognising what the subject actually is, not an approximation, so it simply won’t get distracted by anything else in the frame. What’s also noteworthy is that it’s analysing the whole frame, all 20.4 megapixels at a time.
The aerobatic aircraft proved more of a challenge because of its sheer speed and that, at altitude, it presented a very small target indeed. Nevertheless, the E-M1X’s intelligent tracking hit it more than it missed it which is the opposite of what would have happened with any other autofocusing system, mirrorless or D-SLR. Analysing the frames later revealed an almost perfect sharpness strike rate with the rally cars and a much higher percentage of ‘keepers’ with the aerobatic aircraft than would have ever been the case shooting with anything else.
When it first introduced the multi-shot high-resolution capture function in 2015 on the E-M5 Mark II, Olympus said at the time it would eventually be able to make the facility available when shooting hand-held so you wouldn’t always be restricted to using a tripod. It was a bold promise, but a High-Res Handheld mode is indeed now available on the E-M1X, enabling either RAW or JPEG images with a resolution of 50 megapixels (versus 80 MP in Tripod mode). Depending on how the camera is being held, up to 16 images are captured and then combined into the one high-res frame. Thanks to the camera’s immense processing power, image stabilisation is being switch on and off, frame by frame, during the process, ensuring sharp images. The pixel shifting is ten performed mathematically – rather than physically – after all the images in the sequence have been analysed. Clever or what?
The new ‘Live ND’ function also works by capturing multiple exposures, but with the image stabilisation running continuously through the process… so, again, each frame is sharp. The number of captured frames varies according to the Live ND setting – which ranges from ND2 to ND32 – with a progressively greater number of short exposures being captured to create the motion blur that would be the result of using a conventional neutral density filter on the lens. The benefit is in bright conditions where you can have the effect of a longer exposure time when still using low ISO settings and wider apertures.
A newly-designed gyro sensor – developed to specifically meet Olympus’s requirements – delivers the increased effectiveness of the E-M1X’s in-body image stabilisation which extends the amount of correction for camera shake to seven stops (and up to 7.5 stops with the M.Zuiko Digital ED 12-100mm f4.0 IS PRO zoom). Again, there are clear practical benefits to be derived from this technology, namely yet great scope for hand-held shooting, even with longer lenses.
With a great many sports and action shooters still using D-SLR systems, this market remains largely unconquered by mirrorless cameras, but the OM-D E-M1X represents the most compelling argument for change yet.
For more information visit http://www.olympus.com.au