ARTICLE

The evolution of Canon's Eye Detection AF technology

Discover how Canon developed autofocus technology capable of tracking the eyes of birds in flight, and how this has transformed photography in the field for bird photographer Jonas Classon.
A common goldeneye swimming towards the camera, reflected in the water below.

It is particularly difficult to get the eyes sharp when a common goldeneye swims straight towards the camera. Canon's Eye Detection AF handles the situation with ease. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens and a Canon Extender EF 1.4x III at 1/2000 sec, f/5.6 and ISO2500. © Jonas Classon

Getting the eyes sharp is essential in portrait photography, but have you ever stopped to consider how important it is even in bird photography?

Sometimes the two genres of photography can have very similar aims. "I try to find the character of the bird," explains specialist bird photographer Jonas Classon. "I really try to go inside. Sometimes I feel a strong connection with a bird, and when I'm able to frame it and show this to my viewers, that's what it's really all about for me – to really feel that connection with a wild animal."

Beyond that, imagine the difference it must make to have an autofocus (AF) system capable of detecting and locking on to a bird in flight, even in the most challenging conditions, such as low light and obstacles constantly getting in the way. "It took me 10 years to be able to take action shots of Grey Owls, with a lot of practising and a lot of missed shots," says Jonas. "As soon as I learned how, the Canon EOS R5 and Eye Detection AF came along.

"Yesterday evening, an hour after sunset, in the forest, I took a photo of a Great Grey Owl hunting. I couldn't do that a couple of years ago.

"When Canon's animal eye detection AF came in, everything changed. It enables me to shoot for an extra hour, and for me that is when all the action is."
A Great Grey Owl flies towards the camera, with its wings blurred by motion and the foreground greenery and background trees out of focus.

As a Great Grey Owl flies towards the camera, the Eye Detection AF is able to keep the eyes pin-sharp while the wide aperture softens the background and foreground. As well as isolating his subject in this way, the wide aperture setting enables Jonas to shoot longer into the evening, when owls are most active – and it's all made possible by the AF taking care of focusing. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens at 1/1250 sec, f/4 and ISO4000. © Jonas Classon

A common goldeneye launching itself out of the water, kicking up a row of splashes across the frame.

With the AF effective from edge to edge, photographers don't need to change focus points if they want to recompose during an action scene. Jonas has found that when the AF is locked on to a bird’s eye, it just keeps tracking even when the subject moves to the side of the frame. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens and a Canon Extender EF 1.4x III at 1/1600 sec, f/5.6 and ISO1250. © Jonas Classon

The development of animal Eye Detection AF

Modern digital cameras feature autofocus that can not only automatically track moving subjects, but even track those that move at great speed. Canon has continued to push the boundaries, developing AF technologies that are not only capable of tracking faster movement but also a greater variety of subjects, including dogs, cats and birds. But it didn't stop there – Canon strove to develop AF that can remain focused on subjects as small as a bird's eye.

The latest iteration of Canon's EOS iTR AFX system, built on deep-learning artificial intelligence (AI), enables the latest cameras including the Canon EOS R3, EOS R5 and EOS R6 to recognise and track subjects with new levels of accuracy. The technology grows out of years of development.

Hear more about the way Deep Learning is changing photography in this episode of Canon's Shutter Stories podcast:

In order to develop AF capable of tracking the eyes of dogs, cats and birds, the AI first had to be trained to recognise such subjects. Even among fast-moving subjects, birds are uniquely difficult – there are so many species, with so many differences in shape and size. Swans, for example, are large and white, with long necks and relatively small heads, while parakeets are small and brightly coloured, with short necks and relatively huge heads. Finches have small, pointed bills, while toucans have enormous, colourful beaks. Some birds have prominent crests or combs on their heads, while many have a smooth, rounded crown. Hummingbirds are tiny enough to fit in the palm of a human hand, while several species of albatross can have wingspans of more than three metres.

By comparison to the differences between an ostrich, an owl and a penguin, the variations between species of dog, even those as diverse as Chihuahuas and Great Danes, are minimal. Not only that, but the form and posture of birds can vary radically, from resting on a branch to gliding with outstretched wings.
A technician wearing white gloves cleans the sensor of a Canon camera.

Do you own Canon kit?

Register your kit to access free expert advice, equipment servicing, inspirational events and exclusive special offers with Canon Professional Services.
A black-headed gull taking off from water in dawn light.

A black-headed gull takes off in the early morning light. By using the EOS R5’s vari-angle screen and relying on the Eye Detection AF, Jonas was able to put the camera on the ground and concentrate on composing the image. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens at 1/1600 sec, f/4 and ISO2500. © Jonas Classon

A flock of whooper swans flying against a pink sky.

Jonas was tracking a pair of whooper swans when suddenly a larger group appeared. When other birds interrupt a scene, the AF will continue to track the bird that was originally targeted unless the photographer intentionally changes the point of focus. Taken on a Canon EOS R5 with a Canon EF 200-400mm f/4L IS USM Extender 1.4x lens at 1/500 sec, f/4 and ISO1250. © Jonas Classon

Every bird, every form

To train the AI to recognise such a difficult subject, a vast amount of data was required. Canon gathered a huge variety of images of numerous species of birds in every imaginable position – resting, in flight, with wings spread for take-off, wings flapping, and wings folded in a dive. Then, it was necessary to teach the AI what was not a bird – during development, all manner of subjects were mistakenly identified as birds, including flowers, letters on signs, and outstretched human hands.

Step by step, mistakes and misidentifications were corrected, painstakingly building an accurate system with an outstanding hit rate. Of course, that wasn't the end of the development process. The next step was to refine the quality of AF detection in real-world use.
To train their AI, Canon's development engineers chose Kakegawa Kachouen bird sanctuary in Shizuoka Prefecture, southwest of Tokyo, as their primary testing site. The sanctuary is home to an incredible variety of birds including cranes, swans, penguins, peafowl and even majestic birds of prey such as hawks and owls, which are free to fly when and wherever they please, making the sanctuary an ideal environment for testing bird photography.

It proved even more difficult than the engineers expected. On top of all the challenges we've mentioned, birds won't simply stop and pose for photographs. Even if the AI had become adept at recognising still images of birds, it had to learn to identify them in motion and track them as they continued to move, with their shape transforming as they flapped their wings, for example. This was critical if the AF was to function for video capture as well as stills.

After many long months, Canon's developers were confident that the system could deliver AF performance of the highest quality, even with the most difficult subjects in the most challenging conditions.
A groom and five groomsmen of various ages, most wearing sunglasses, walk towards the camera smiling broadly.

How Canon's intelligent autofocus works

Find out how the intelligent autofocus system in the Canon's latest cameras "makes it practically impossible to miss a shot".
A pair of great crested grebes nuzzling beaks in the water.

When great crested grebes perform a mating dance, their heads spin quickly from side to side. Before Eye Detection AF, Jonas says, he would get just two or three sharp images in a series of 10. Now his hit rate is close to 100%. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens and a Canon Extender EF 1.4x III at 1/500 sec, f/5.6 and ISO1600. © Jonas Classon

A black tern flying directly upwards towards an orange insect.

A black tern locks its eyes on an insect in flight. Canon's AF tracking is able to cope with the quick movements of the bird leading up to the strike. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens at 1/3200 sec, f/4.5 and ISO1250. © Jonas Classon

Eye contact

"The very first time we tested Eye Detection AF with photographers, we went to a safari park," says Canon Europe Senior Product Specialist Mike Burnhill. "On the way in, there was a black bird splashing about in a puddle. We pointed the camera at it, and it locked on to the black eye of a black bird in muddy water. It was amazing. The photographers were sold straight away."

Animal eye detection and tracking AF is only one part of Canon's intelligent autofocus system, which delivers a whole range of benefits including outstanding low-light AF performance and class-leading AF acquisition speeds. Thanks to Canon's Dual Pixel CMOS AF II technology, every pixel in the sensor can be used for both imaging and AF, which means the AF can acquire and track subjects across the entire frame, despite obstacles and distractions. "Because every pixel can both focus and take an image, they don't have to be one or the other," says Mike. "They are both. That's the magic thing. And they provide so much data, it gives us the ability to do all kinds of stuff, like focus right to the edge." The processor scans the entire scene up to 120 times per second, evaluating out-of-focus areas as well as in-focus areas, anticipating when objects might cross the path of the subject in order to keep tracking the subject without interruption.

The result is autofocus capable of detecting the eye of a bird in flight and remaining locked on to it.
A bird with a black-coloured head flying towards the camera, strikingly sharp against a blurred background.

Jonas was delighted to discover how effective the Eye Detection AF could be, even in poor light and with a distracting background. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens at 1/1600 sec, f/4 and ISO3200. © Jonas Classon

A black tern swooping low to the ground through plant stalks.

A black tern searching for insects in the reeds. Although Jonas took the shot through the reeds, the Eye Detection AF managed to keep the bird in sharp focus. Taken on a Canon EOS R5 with a Canon EF 600mm f/4L IS III USM lens at 1/1000 sec, f/4 and ISO4000. © Jonas Classon

Game-changing

For Jonas, out in the field, this radical leap forward has been game-changing.

"Before, I was stopping down to f/5.6 or f/6.3 just to have some extra depth of field to get the focus right. Now, with Eye Detection AF, I can really open up the aperture," he says. "I can go down to f/4 or whatever my lens is able to do. Not only does it enable me to shoot for an hour longer, it means I can get really nice backgrounds and foregrounds. It opens up a lot of things."

Animal Eye Detection AF has also revolutionised where Jonas can photograph. Before, he would tend to limit himself to open fields where he could get a clear shot, but now he goes wherever the owls go. "Now, I can take more creative shots by including foreground grass and other natural elements, because the autofocus will have the bird covered," he says. "That's a big difference for me, because now I can take action shots of the owls hunting in the deep forest without the trees interrupting the autofocus. That gives me opportunities."

Finally, Eye Detection AF has reliably delivered remarkable results. "These days, it's more rare to have shots end up out of focus," Jonas says. "If I have an owl flying straight towards me, I'll have 18 out of 20 pictures tack-sharp. It means I can select the perfect angles of the wings and the best composition."

Jonas says Canon's animal Eye Detection AF opens up new possibilities for him, enabling him to shoot almost without limits, and he still finds it astonishing.

"How can you find a black eye on a black bird, on a cloudy day, with distracting backgrounds?" Jonas asks. "How is it possible? At first, I didn't believe it. But Canon has nailed it. It's amazing."

Написано от Mark Alexander and Alex Summersby


Related articles

Get the newsletter

Click here to get inspiring stories and exciting news from Canon Europe Pro