Samsung Gear360 VR camera experiences

Hi, are there anybody, using Samsung Gear360 camera?

I bought one, and took cca 50 000 photos with it already. The camera is mounted on my bicycle helmet, and I’m tracking cicleways on the hills, and other highways in the forest, in the city, etc…
About the softwares I’m using: I don’t like the original Gear360 app, because it is painfully slow - it last cca 10 sec to take a photo on my Galaxy S7 phone. The other disadvantage: tapping the link on the phone screen -mounted on my bike is difficult,
I bought a “Gear 360 Value Kit” because in it there are a remote control for the cam, with it I can take a photo in every 1 second. This remote control is perfect on the brake handle.
On the terrain I’m using the camera controlled by remote control, and the phone only is recording the track gpx file. In the cam there are not gps and compass, so at home I correlate the photos to the track with the JOSM free software.
On the go there is no need for samsung phone, I take the photos without phone, and II’m recording the track gpx file with my old LG G4 phone. At home stitching and saving the photos I’m using the S7 edge phone.
The cam exposition is fast, I often take the photos at the speed 15 km/h.
To set the directions of the photos I’m using the free “interpolate_direction” python script, and JOSM “PhotoAdjust” plugin, or the mapillary plugin.
The cam battery can work about 3 hour, taking hundreds of photos.
I bought 2 spare battery for the cam…
The remote control battery is yet good after 50 000 pushing.
In the cam there are a perfect giroscope, the photos are always correctly horisontal.
Here is a sample: (Lackenhof- Lunzer see - Waidhofen) There were rainy and sunny weather too.
The pictures were taken at the speed of my bike: 0-20 km/h.
On my photos always are visible my blue helmet and the Gear360 logo

Are there any better idea, to accelerate the workflow? :slight_smile:

Thank you

3 Likes

I see the camera has a gyro to ensure the photo is always level. That must make it possible to mount on your helmet, something I can’t really do with my LG360.

How do you find it handles tilted heads, turning you head around, looking down etc? Any samples you can link to?

I’m used to hold my head in position when taking photos. But if there are something interesting for the edithing OSM map, then I turn my head to the direction of the object, and I’m correcting the direction when upload the photos. (JOSM mapillary plugin, or PC mapillary uploader webpage)

I’v made a trial:
the gear360 cam tilted 45 grad, I made a photo, saved to my phone, then uploaded to google photos cloud.
Opening on PC in google photos, the photo perfectly horizontal - so the giroscope is perfect.

Can you give a link to a uploaded mapillary sequence? I am interested to see the quality.

Here is a sample: Lackenhof - Lunzer see - Waidhofen
On my photos always visible my blue helmet and the Gear360 logo

Hi, I’ve just got my first test runs published in Mapillary using Samsung Gear 360 2016-model, example here: Mapillary

A simple setup on my cars back window is suitable for up to 60-80km/h speeds at least. 2016-model has lower quality accelerometer, images are always tilting one way or the other when accelerating. At least the still image resolution is much higher :smile:

Car roof takes quite high amount of image space even with this setup where camera about 25cm above roof, so this is definitely a thing to improve. Antenna pole has M6 thread, could be used for securing… Roof racks could be simple solution too.

Camera is triggered manually by standard Samsung bluetooth remote from “Gear 360 Accessory Value Kit”. A hacked intervalometer-version is under construction. A standard SLR intervalometer should trigger the switch, but the challenge is the miniature size of the switch and how to solder it. Copper pins could offer a solution where one wouldn’t have to tamper with the original switch at all, but this is something I need to ask help from some electronic wizards at the local hacklab :smile:

1 Like

Geotagging the photos and set the directions:
I’m recording the tracks with locus on my phone, then correlating the photos with the recorded gpx tracks in the JOSM free software - so the photos got the geotags. Then I’m running the free python script: “interpolate_direction.py”, this set the direction of the photos.
This whole process last cca 5 min on my pc for 1000 photos.
When I’m beginning taking photos, very important to set the exact time on the gear360 - so I turn on the gear 360 app on my phone for 10 sec, then turn of. Afther the process the time on the 2 gadget is the very same. After changing battery in the cam, I have to set the exact time again.
If I forgot setting the exact time before taking photos -the gear 360 cam often take 30-60 sec time error - so in the correlating process I have to set time offset manually. It is unsteady and taking many time to set the exact offset.
When I’m taking the first photo on my track - I’m taking the frist photo allways is near a spot - which is visible on the map - so after geotagging I can supervise the exact place of the photo. Taking the first photos very importan driving fast - so I can see better the exact geotagging.

I just bought a Gear360: the 2016 version’s price was dropping so I was glad I could make photos for mapping. However I couldn’t imagine that software support is so poor!

I would really appreciate if somebody could help me how to stitch photos properly. I don’t own a Samsung phone, I use this port of the official app: […] It works, but it’s slow and I guess it doesn’t have all functions (for example I can’t configure the time on the device…)

I also use Linux, and using ActionDirector with wine didn’t work for me. I tried to use Hugin with and without this gear360pano script: […] but I didn’t get an acceptable result. (Running from terminal I got an error of enblend; applying the script directly in Hugin GUI gives a photo with a heavily jagged line between the stitched images.)

I also tried to simply upload the original photo to mapillary: Mapillary It’s not well stitched.

I found this online tool: [nadirpatch] Here you can upload a photo and download the stitched one after. The size of the photo was doubled at the end (20 Mb), however uploading this picture to mapillary gave a much better result: Mapillary Of course that’s not a final solution if I would like to make hundreds of photos…

Couldn’t Mapillary give some support for this device? In my opinion the hardware has pretty good quality and as the new model came out, the 2016 model became much more affordable, maybe many people will have it. If Mapillary at least at the backend side could stitch the pictures just like nadirpatch did, it’d be a huge benefit.

Any help is appreciated: do you know a better solution to use the camera on Android? Does somebody have experience with wine and ActionDirector? Maybe I miss something basic in Hugin?

@sct @Tamas how were you able to use the Gear360 logo on your helmet? Value kit is out of stock: is there any other remote which could work with Gear360?

Thanks in advance!

(I had to remove the links because I was not allowed to post more than 2 as a new user…)

1 Like

I think in the new gear360 android app the gear360 logo is missing already. The old app was better for me!

Any news on this?
The 2016 Samsung Gear 360 really seems like a really good match for Mapillary except for this interval photo issue.

Is manually pressing the bluetooth remote really the only solution? Did someone find a way to create an app that sends the right bluetooth commands?

It’s strange that the (seemingly) official Mapillary 360 page says the interval is 0.5 seconds and resolution 25.9 Mpx. It seems by default 0.5 s is only for the video mode so either that’s misleading or they know something we don’t.

I started a thread about this particular camera back in May last year.

The camera has been hacked - you can get a root shell. That could enable a feature purely in software where the camera takes pictures at short intervals.

Check out this github project, including the “issues”: GitHub - ultramango/gear360reveng: Some internal information on Samsung Gear 360