[Hardware] Canon EOS R5 II in-depth review: Digital Photography Review

[Aronus]

The biggest change in the EOS R5 II is the move to a Stacked CMOS sensor. As we've seen before, this allows the camera to read out its sensor much, much faster than conventional BSI or FSI chips. This speed allows the camera to capture 8K video at up to 60p and bursts of stills at up to 30fps.

If 30fps is too much for the shooting you do, the camera can be customized to shoot at 20, 15, 10 or 5fps instead. We found the readout rate in stills mode to be 6.3ms (∼1/160 sec) and that it appears to read out in 16-line chunks. This is a fair bit slower than the likes of the Z8 (∼1/270).

The EOS R5 II's sensor uses Canon's Dual Pixel AF design, with two photodiodes for each pixel location, meaning almost the entire sensor can be used for phase-detection autofocus. But it does not have the cross-type arrangement used in the EOS R1. It also omits the original R5's Dual Pixel RAW option, but we suspect very few people found a good use for this.

The EOS R5 II gets the same updated version of Eye-control AF that appears in the R1. Like the first modern version in the EOS R3, this lets you select a subject to focus on by looking in its direction: as soon as you half-press the shutter button or hit AF-On to initiate focusing, the camera will pick the subject nearest to where it thinks you're looking.

The latest version has sensors with a wider field of view, to get a better view of the eye as a whole, and revised algorithms that should work better when the eye is further from the viewfinder. This should be particularly useful for people wearing glasses and is supported with an eyeglasses detection mode, to avoid it getting confused by the additional lens and distance between the sensors and the photographer's eye.

The EOS R5 II has the same 'Digic Accelerator' co-processor used in the R1. The way Canon talks about it is directly comparable to the way Sony describes its 'AI processing unit': a dedicated processor focused on crunching its way through the complex algorithms generated by AI mechanisms, which means the main processor can stay focused on things like distance measurement and communicating with the lens, rather than trying to do everything at once.

Canon says this processor helps run more complex exposure and white balance algorithms, too, boosting the camera's fundamental functions, as well as powering some of its novel features.

As with the R1, the R5 II can be set to capture up to 1/2 a second of action before you fully depress the shutter, so long as you have the button half-pressed beforehand. It can also pre-capture 3 or 5 seconds of video.

One of the biggest features the R5 II shares with the R1 is its slate of Action Priority AF modes. These are based on machine learning trained on how different sports look: how active players' bodies are positioned and what the ball looks like at key moments. The camera is able to follow the ball and work out which player to prioritize. For instance, Canon says the Football/Soccer mode has been trained to recognize the following actions occurring: "Shoot, header, short pass, long pass, dribble, clear, place kick, save by keeper, throw in, sliding tackle".

At launch, the EOS R5 II will have AF modes for Football ('Soccer' on cameras with US firmware), Basketball and Volleyball. Canon did not say whether it plans to add other sports, via firmware, later, but implied it would be possible.

In addition to the generic subject types the camera has been trained to recognize, it's also possible to register up to ten sets of ten people in the camera's memory. If engaged, the camera will try to give priority to those individuals if it identifies them in a scene.

This could be used to ensure you get photos of specific players at a sporting event or, for instance, during wedding shoots, where you can set the camera to prioritize the betrothed, rather than risking the camera getting distracted by other members of the wedding party that happen to get closer to your AF point.

You can even create a priority rank of the people you've identified so that the camera knows to focus on the bride ahead of the groom, ahead of the maid of honor, and so forth. Despite giving the camera only a single image of each subject, it proved pretty good at recognizing the people we asked it to prioritize.

Like the R1, the R5 II gains two post-shot options to reprocess your files using the full power of its AI-derived algorithms. It's not got the processing power or battery life to apply these as you shoot, so you'll need to select which images to process, after the fact. Even with a dedicated 'accelerator' chip, both processes take several seconds to apply to each file.

The first option is to up-res the images to twice their original resolution (four times the pixel count). This uses machine learning to anticipate what might occur between the captured pixels, to boost the apparent resolution. Somewhat unexpectedly, this process is conducted on JPEG or HEIF images only, rather than the Raw data.

The other option is to apply complex "neural network" noise reduction to files. This again is based on machine learning and attempts to distinguish between noise and detail to give a clean but detailed image. The option can be applied to Raw files but Canon says that the JPEGs that are then created cannot then be up-sized.

 

Canon EOS R5 II in-depth review: Digital Photography Review