Over the past few years, I've been shooting video from my bicycle. I'm taking a first look here at the capabilities of the Garmin VIRB 360 camera as used on a bicycle.
I thank Robert Efthimos for permission to use the video embedded below, which he shot using a Garmin VIRB 360 camera. For the best results, expand the image to full-screen and click on the gear-wheel icon at the lower right to choose the highest resolution which plays smoothly. Once the image is in motion, you may drag your computer's mouse pointer inside it to look in any direction.
The VIRB 360 actually is two cameras in one, facing front and rear. Each has a super-wide-angle fisheye lens with a viewing angle of more than 180 degrees, producing a circular image with the usual fisheye distortions: straight lines near the edges curve inward like parentheses. The front and rear images together cover all directions. Software corrects the geometry and stitches the images together. As the example above shows, YouTube supports dragging the image to look in any direction, even directly overhead, though only in one direction at a time.
The image below is a downward view, to help you understand how the camera works. You are looking at the tail of Robert's bicycle helmet. His back is at the bottom of the picture: the scorpion logo is on the back of his Castelli jersey. The camera is only a couple of inches above the top of the helmet, and so his back looks small compared with the helmet. There is a blurry transition line just above the middle of the image, where the forward part of the helmet looks wider than the rearward part. This results from the forward-facing lens being farther forward than the rearward-facing lens, capturing a wider part of the helmet. The bicycle at the right in the photo is far enough away that the difference in position of the two lenses doesn't matter. The camera gets away with this for objects more than a few feet away because it is small. The camera's attachment to the helmet is in the area which is not shown, between the front and rear images. The white circle in the middle of the image points directly downward, and the arrow above the letter "N" indicates the direction of forward travel.
The camera has internal image stabilization. Garmin describes three image-stabilization modes:
Robert's video was shot in follow mode. The helmet can be seen bobbling around, while the surrounding scene remains steady. If you point the view downward, the white circle changes position on the helmet as Robert rides over bumps in the road.
More remarkably, as shown in the image above, the horizon remains level despite the tilt of the helmet as Robert leans to turn. (You may see this moment in motion at 31:30 in the video). The view remains upright even when the horizon cannot be identified. That wouldn't be possible with image stabilization that responds only to the content of the image. Neither is the camera identifying the horizon according to the direction in which gravity seems to pull: that direction changes when the bicycle turns, speeds up or slows down. The image stabilization is, then, responding to multiple inputs including the camera's GPS. GPS does not identify the location and orientation continuously, and so the camera also has an inertial sensor to adjust orientation between GPS readings. I suspect, though, that the image stabilization would degrade indoors, or in a tunnel or concrete canyon where GPS signals are not available.
The 360-degree image has no edges, and so it solves another problem with image stabilization: Even if the camera's direction of view changes wildly, there is no need to crop the image so its edges don't show.
The camera's GPS maps the bicycle's route. In the post-processed computer display above, the route is visible as a jagged line, with the white dot indicating the location at the time shown in the video image. The cyclists were at a turnaround where they arrived and departed on the same road.
The two semicircular displays at the bottom of the screen report elevation and speed. A circular display, directly overhead in the image, indicates the compass direction of travel.
Optical image stabilization is difficult in a small camera with a wide field of view. The camera uses digital electronic image stabilization, and so image quality will degrade in dim light, when a longer exposure time is needed for each frame of video. In the nighttime clip below, shot by Philip Carlson, the camera is in follow mode. The data display in the image turns with his head, but the camera image remains oriented in the direction he is traveling, and blurs as the sensor rotates. With more conventioanl cameras, the image blurs but also changes direction, a more familiar effect.
The VIRB 360 is a technological marvel. It is excellent to document a bicycle race for review, because it shows all the other racers surrounding the one who has the camera. It must be mounted on top of the helmet to have an open view in every direction. The single camera avoids the need to synchronize front- and rear-facing cameras in post production.
There are also disadvantages.This camera is advertised as rainproof, and waterproof down to a depth of several meters. But as with most action cameras, it will not take good video with raindrops on the lenses. Treatments such as RainX, which spread out individual raindrops into a thin sheet of water, may help, though I'm not sure whether they are compatible with the lenses' anti-reflection coating. A waterproof shell for the camera is conceivable, with its seam along the line that is not visible to either lens, and would reduce the problem with raindrops, but the manufacturer does not offer such an accessory.
Without a shell, the lenses are exposed and vulnerable to damage by sand and grit in the air. This is especially so with action sports such as bicycling and motorcycling where the camera is moving forward at speed. The lenses protrude, as they must for a 180 degree view, and so also they are vulnerable if the camera is dropped. However, the outer layer of glass of the lenses is easily replaced.
Underwater, the image will go out of focus, and 3D will not stitch, due to the refractive property of water. Underwater cameras have either a specially-designed underwater lens or a flat window. A flat window is not possible for a lens with a field of view extending beyond 180 degrees.
From its high perch on top of the helmet, the camera doesn't offer much of a view of the cyclist carrying it. Mounted on a stalk extending upward and forward -- or rearward -- from the helmet, the camera could show the cyclist as well as a forward and rearward view, but it would unbalance the helmet. This camera is rather heavy compared with other action cameras, worsening that problem. Mounted on the handlebar, it could provide a good view to the front and a complete if odd view of the cyclist. Another option that provides a a 360-degree view and a view of the cyclist is to mount the camera on a pole behind the cyclist.
Pairing these cameras for a 3D view is theoretically possible, but the 3D would work only for a view to the front or rear. Each camera would show in the other's image and would move around with the image stabilization. The manufacturer does not offer a 3D option.
In the Garmin's stabilize mode, the view to the rear, as well as the front, changes as the cyclist's head turns. When the cyclist looks to the left, the rear view points uselessly off to the side of the road. But in follow mode, the camera points in the direction of travel regardless of head turns.
I haven't sprung for a VIRB 360, and so I use two cameras, for a front and rear view. I like to mount my front-facing camera on my helmet, so it points in the direction where I am looking, but I mount my rear-facing camera on the rear rack of my bicycle so it is always showing the road behind me. I'll show only the front view much of the time, but as needed, I'll show the rear view as a picture in picture, even zoom in on it so it fills the entire video frame.
Because the eyes, not only the head, turn when looking over the shoulder, my helmet-mounted, front-facing camera often doesn't turn far enough to show what I am looking at. This issue can be addressed with the VIRB by panning the view in post production, but it could be useful if the camera had a mode to automatically rotate the view farther than the head.
Both improvements I've suggested are theoretically possible in software, and I'd hope that Garmin implements them. They may be available in aftermarket software packages.
The camera could avoid blur in low light by being mounted on a gimbal, but gimbals now becoming available for action cameras are a bit bulky and would obstruct part of the 360 degree view of this camera.
The camera has wireless connectivity to a large variety of devices, allowing remote control, viewing and data input/output. Garmin mentions surround audio with four microphones in the camera. That's nice but they pick up a lot of wind noise. At the expense of additional complication, the camera can connect to Bluetooth microphones shielded from the wind.
The manufacturer claims that the camera can run for an hour on a battery charge. This is not long enough for many bicycle rides, though an external power source may be used.
As of July, 2017, the VIRB 360 is expensive at $800 US. On the other hand, it substitutes for two cameras.
A good review of the VIRB 360 covering some additional topics is available on YouTube, and there is an exhaustive review online here.