Buying a new 360° camera for mapillary capture

Hello, I would like to buy a 360 camera for work. the objective is to capture all the streets of the city mainly by car and then put them under mapillary.

I hesitate between the virb 360 garmin and the Mxx gopro or fusion.

what do you suggest?

1 Like

if you’ve already narrowed down the list to these 3, you are probably better off with a fusion. it’s now discounted and gets the best image out of all 3 of those

1 Like

Hello, thank you for your opinion.

it seems to me that these 3 references are the most adapted to our needs but I remain open to other brands of versatile 360 camera

it depends on your budget and workflow a lot. If you can afford to, imagery from insta360 pro would have more detail, and there is a gps module. If you are ok with matching the images to gpx tracks post-capture, there are cameras with decent image quality (obsidian go, aleta sc) which are a bit cheaper than the insta360.
When I need slightly better image detail without necessarily seamless 360 (mapillary has combined panning so not much of an issue), I use a 4xgopro hero 5 rig. There is more post processing since it doesn’t record image heading, but otherwise comparable to the fusion with better image fidelity

thank you for making me discover all these different brands.
we want to have a large production capacity (city of + 80,000 people) and the simplest possible processing line. Indeed, we want to integrate the whole city on mapillary.

the insta360 pro 1 and 2, which I am discovering, are impressive and very complete.

is post-processing and geolocation for obsidian easy?

Take a look at this thread about the YI360:
http://mapillary.trydiscourse.com/t/yi-360-vr-camera-review/2935

I am working on the “perfect workflow” to do what you are planing to do (I think). I am almost there… but there is a snag now that since the iOS update (to version 13) their app can no longer work with my iPhone.

When you are planing on doing it by car, do keep in mind the interval at which you are going to be able to take images!

When your camera can do one images per 2 seconds and want an ideal 2.5 meters between images your maximum speed is 18 km/h (speed: fast cycling) Driving a car will easily go to 50 km/h, resulting in 7 meters between images (upwards)… not ideal… When in “MP4 movie mode” one can have 24 frames (read: images) per second and drive up to a whopping 216 km/h and still have 2.5 meters between images!

I have built my processing software to handle MP4 files which creates JPG’s from the movie, match GPX (in sub seconds!) removes images when the distance between the previous is less then X meters, uses Hugin to do a (near) perfect stitch and a whole bunch of other stuff I ran across…

I am working on a project where I want to map my (rural) village (about 11km2) all in 360 images, ranging from road, cyling road to forest path. This year I have been working on “the perfect workflow”. Starting with getting the images as good as possible (ranging form good stitch work, image settings, best moment to take the images, setting it all up right every time, best GPS hardware, researching (dis)advantages of walking, cycling and going by car, etc etc) so that I can do the work next spring in a relatively short period and let the computer do most/all the work :wink: In the meantime I’m working on a platform where my neighbor (who has the village archive) can enter/process the village history… going to combine with public (geo) data etc. etc… to create an intuitive and interactive platform to unlock and show the village history (when my wife allows me to and the work for my own ICT company lets me :stuck_out_tongue: )

I think there is quite a bit of overlapping in our goals. In my decision making process I saw the three cameras you mentioned. In my opinion the YI360 is better, only the software and external GPS are the downsides I am working on (with that program I am creating). The insta360 pro cameras are definitely superior to all cameras mentioned. But quite a bit over budget for my project so I did not consider them. At what speed and quality can they record movies, or take images? (And do you need that extra quality?)

If you want we could call? Do keep in mind I’m Dutch (UTC+1) and you are also UTC+1 when I did my research correctly :wink:

Hello,

thank you for all your details.

To start our work we will choose Gopro FUSION 360 for its speed of implementation, its versatility and its adaptation to mapillary.
Also, the new gopro MAx is more suitable and more efficient than fusion for our use?

We will see in 1 year if we switch to more professional equipment such as Insta360.

Do read this topic: ‘Gopro Fusion Workflow
There are some tips about slow data transfer rates (use card reader) and take note of overheating problems!
http://mapillary.trydiscourse.com/t/gopro-fusion-workflow/1679/56

PS: I think the Fusion is better suited for Mapillary then the MAX (read this topic)

1 Like

probably a good choice overall, if Pros are out of your budget.

If you can get the Max for roughly the same price as the Fusion in your area, with an option to return, I’d try it out - considering you want to cover a large area and simple workflow, the protective lenses and one sd card instead of 2 might outweigh the drawbacks. The main thing to check for would be whether it writes gps data into 360 imagery exif, as the answer I’ve been getting so far was No.

Also, just general points on the Fusion and 360 imagery:

  • Check that you have enough desktop processing power for stitching
  • Consider the mounting options and how hard you are going to use them - those lenses are very fragile
  • Have a look at some of the Fusion imagery already on mapillary and elsewhere, to avoid false hopes in terms of the level of detail. It is generally quite low

I find it worrying that gopro remove the gopro fusion from its sales catalogue. this means that updates will have to stop in the future and all resources on the gopro MAx !

I looked at some captures with gopro fusion and relatively low image quality indeed!!

it was made clear back when the 2.0 beta fw was released for the Fusion that no further meaningful support will be given. gopro can’t really afford it, and they will focus on pushing the max.

low image quality is the bane of 360 capture with consumer cameras, hence why I suggested the more expensive options to you, as well as the multi-gopro rig

Thank you.

Indeed, the multi-device solution seems to combine both image quality and controlled price.
On a multi gopro and in the case of image capture mainly in the car, the processing and worflow for integration on mapillary use which software and can synchronizations be automated?

If Pros are out of your budget, I will consider Mi Sphere 360

Resolution is higher than your other options. Take a look at a sample

1 Like

hi,

with the MI Sphere 360 , what do you use for upload the images ?

It would depend on whether you want the pano image to be seamless on upload or let Mapillary do combined panning between 4 images.
If you want a seamless file to upload, this would require extra processing (likely hugin or ptgui or similar) to stitch the 4 images. This would of course take time.
The heros also don’t have compasses (I think tomtom bandit and garmin virbs do, but I never proper compass data from my virb), so you’d need to first normalise the cameras’ heading in the direction of travel, and then set an offset for each based on how you mount them. This can be done with mapillary scrips, external tools, and in the mapillary web interface after upload.

Not sure if the mi sphere would work for you, as it is slow, doesn’t have built in gps and the resolution is not much better than fusion

Hello, what is the work flow for? Professional camera are not a insta360, there are systems meant directly for city asset extraction I’d consider as an upgrade. I’m buying the GoPro Max today to pilot my project but would probably upgrade to a mobility mapping system if the clients want better quality photo and GPS/IMU.
Single high end camera (like DSLR) give the best image, but make sure you get a CMOS, mechanical shutter on a pro system for higher movement speed. Check out the lady bug stats on flies website for an older camera. Millions of miles recorded with that one.

I have done one test with my YI360 on the roof of my car:

In my opinion the quality is quite nice. But I didn’t like it that my car took so much space in the image. This could partially be fixed by mounting the camera higher. But I decided that I will work on a setup mounted on my bike, since my bicycle is obviously quite a bit smaler then my car is :wink:

These images are taken in “movie mode”, thus the distance between the images is just a meter or two! Do keep in mind that an (extreme) higher resolution in images almost certainly means a greater distance between the images! What do you need? Think about things like priorities in: good coverage, image quality, fast results, detail, high speeds, countryside, city, accessible (only) by car, scooter, bicycle?

PS: This test was stitched with the software of YI, thus you do see “stitching echos” here and there. This has been solved by using Hugin in my workflow now (much better stitching quality and much more easy implementation in my goal to completely automate my workflow).

Ladybug and the like are overpriced for the image quality their offer, unless a better gnss unit in the camera is needed. The OP specifically asked for an easier workflow

Sorry for my lack in knowledge… I know what you mean with gnss, but what is “Ladybug” (a camera brand?) and “OP” (you mean the original topic starter)?

1 Like

In a linux box:

  • Georeference from gpx file with JOSM.
  • Stich with hugin_executor.
  • Upload with mapillary_tools.