Nice new 360 Camera for Mapillary: PANOX V2

1

Hi folks, I did a quick test of a new 360 camera which looks to be quite good for Mapillary, the PANOX V2 (thank you LabPano for the sample!). Has anyone tried this camera out? If not, I think it’s worth your while to give it a go.

The things I liked about it:

  • 8K 5 FPS mode (so both higher density and higher resolution than GoPro Max which tops out at 5.7K)
  • Built-in GPS (unlike the Insta360 X3)
  • On device stitching of 360s (so you just drag the files right into the desktop uploader, no other processing needed unlike the Insta360 X3)
  • Relatively inexpensive ($399 on the official site and Amazon, but I’m seeing it as low as $299 on Aliexpress)
  • Standard tripod camera mount (so you don’t have to buy an adapter like you do for GoPro)

Things that could be improved:

  • The UI on the camera feels a bit less polished than GoPro or Insta360
  • Acquiring GPS satellites can take a bit of time (at first I thought it wasn’t working)

I haven’t done extensive testing for long recording sessions or detailed image quality comparisons, but overall this feels like a solid camera. Here’s a sample Mapillary sequence I shot at 8K 5FPS.

What do you think?

003
0031400×1400 38.1 KB

8 Likes
2

The pictures are better than those of the Gopro Max.

3

Looks good!
Some nitpicking:
In low light it seems to increase exposure time a bit too much, leading to motion blur, it should rather increase gain, noise would be more acceptable than motion blur. Mapillary
What exposure did it set here, maybe it can be read from the EXIF?
Might there be a setting to limit exposure time?

Other things:

  • The battery seems to be fixed?
  • The app is not available for my phone. (Pixel 6 Pro with Android 14
    https://play.google.com/store/apps/details?id=com.pi.panox
    From your post I read that it is not necessary at all? Does it contain shady tracking and spyware like too many other Chinese apps?
  • The manufacturer’s page says it has a fan for cooling. Hopefully helps against overheating. But might limit flexibility in rainy weather, or generally in dirty environments.
4

For the time being we have to accept to make only 360s around noon in good weather. Otherwise it is not pleasing to the eye.

2 Likes
5

@enteq - good point about the exposure in that moment. Good find, playing with settings might be an answer, or it could be a limitation. Regarding the other things you mentioned:

  • Yes, the battery is fixed in place (but the usb-c port is exposed, so it should be possible to run it from an external power-bank - though I haven’t tried that yet)
  • That’s strange that the app is not available on your phone. I have a Pixel 7 with Android 14. Maybe its country dependent, what country is your Play Store set to? I’m sure PanoX support could help there. I haven’t tried setting up the camera without the app, but FWIW the company that makes this is LabPano which has been around for 10+ years and made consumer and professional cameras: https://www.labpano.com/en/about
  • Good point, ideally we should avoid capturing Mapillary imagery in the rain, and I haven’t done testing in dirty environments.
6

Hey, looks interesting indeed! :slight_smile:

@boris

  • Can you lock ISO speed and/or exposure time manually? :white_check_mark: full photometric control
  • Is the 5 fps mode in video or image mode? :white_check_mark: video
    • The examples you have provided suffer from ugly red and blue artifacts (hence I’m assuming video mode) on grey surfaces.
  • Can you set time lapse mode in image mode?
    • If so, what is the slowest and fastest fps?
  • Can you manually set the JPEG compression ratio? :white_check_mark: sort of: will export to DNG in the future
  • Can you manually set the video compression ratio?
    • What format does the camera encode video to (h.264, HEVC, etc)? :white_check_mark: h.264 and HEVC
  • Can you set the chroma subsampling factor in image and/or video mode manually?
    • If not, what chroma subsampling factor does the camera output? I hope it is 4:4:4 in image mode. :white_check_mark: sort of: via DNG
  • Is the transparent outer lens cover made out of plastic or glass?

Btw, stitching looks fine to me. :+1: A standard tripod mount is definitely a must on such a camera.

Image mode supports 72 MP, which is finally good enough for Mapillary purposes. @boris Can you capture some 72 MP image mode imagery for Mapillary, possibly in time lapse mode? Thank you!

1 Like
7

Quality of that sample sequence looks comparable to Kandao 8k (which itself is not too different from the 5.7 from Fusion/Max) - there is still a woeful lack of detail in all of these consumer cams, and I wasn’t expecting better given only 2 sensors.

Maybe 72mp still are better (although I have doubts over how quickly it can capture them), but not holding my breath - this is supposed to be the cheaper Labpano camera, and even their existing more expensive models are not blowing Pro/Pro2 out of the water

8

This camera looks interesting! It looks like an insta360 X3 funnily enough but it seems to actually work and stuff. If it works great, then I might get this camera for my next 360 camera upgrade!

1 Like
9

You can set time lapse (called interval) in image mode, but I think it’s not really meant for the Mapillary use case as the intervals as 5, 8,10, 15, 20 seconds. I think you also then need to run a stitch on the camera (where as video is stitched automatically). Video mode is I think what we’d want to use (which offers 8K at 1,2,5 FPS and 5.7K at 1,2,3,4,7 FPS)

Yes, you can set “Middle, High, and Ultra High” VBR

I’m not sure I can tell, I think it might be plastic based on the sound, but hard to tell.

This is called 12K on the camera, here’s an example image: Mapillary - note that this does take quite a while to capture, so again, I would probably stick to the video capture at 8K for the Mapillary use case.

2 Likes
10

Thank you for the additional info. I really appreciate taking your time. Like you have guessed correctly, I was asking because I would favor image mode.

You can set time lapse (called interval) in image mode, but I think it’s not really meant for the Mapillary use case as the intervals as 5, 8,10, 15, 20 seconds. I think you also then need to run a stitch on the camera (where as video is stitched automatically).

No less than 5 seconds is a bummer. Besides, we live in the age of computers and integrated timing circuits. Hence, it is beyond my understanding why software developers force hard coded intervals on users. I am aware that there is certainly a minimum hardware limit but for goodness sake, why do we have to patronize and limit users? I am also pretty sure that both sensors can capture well below a fraction of a second and internal storage can take the data at the required speed.

Anyway, imho in Mapillary’s use case we would need the camera to capture like 1 to 2 fps in image mode and do the stitching after the user ends time lapse (interval) mode. Image mode quality does look better than video mode. You can definitely see fewer digital artifacts and more details. But, I am not sure about sensor noise. Did you set ISO speed 100 manually in the example?

I think you also then need to run a stitch on the camera (where as video is stitched automatically).
…note that this does take quite a while to capture…

Can you also comment on how long does stitching take in image mode? Can you offload or time shift stitching to the smartphone app? Is capturing speed in image mode limited by stitching speed?

Should both sensors be treated as two separate cameras in Android’s Camera API then developing an app for Mapillary’s specific use case should not be that hard. Otherwise, we are left with hoping for the best from LabPano.

Video mode is I think what we’d want to use (which offers 8K at 1,2,5 FPS and 5.7K at 1,2,3,4,7 FPS)

I understand your reasoning but it sounds rather like a workaround for Mapillary’s use case, which oddly enough is basically identical to “Google Street View” mode but apparently LabPano was unable to get it right either. “Google Street View” mode should have been what I have described above. We do not care about instant stitching in this mode. We care about 1. image quality, 2. timing, and 3. GPS data in this mode. Stitching is a post‑processing step in this use case.

Overall, this camera has some great potential but the software needs to be more tailored for Mapillary’s use case. Frankly, I was considering buying this camera (more than any GoPro 360 or Insta) just for Mapillary, which is a huge deal for only one use case. I am not interested in vlogging or the likes. But, image quality alone will not win the game for me here. Workflow and software are important too.

Btw, the example’s time stamp dates Jan 1st, 1970 (the Unix epoch). Do you have any clue how this comes?

11

No, these were fully automatic settings

Stitching is fairly quick in image mode (several seconds?) and you do it async to capture, so I don’t think it has any impact on capturing speed. I didn’t try doing it in the smartphone app.

Yes, I noticed this as well, it looks like the GPS time for this image was not set for whatever reason. I reported it to LabPano (and we’ll also add a workaround to use “created at” time when GPS time is Jan 1, 1970). Note that this wasn’t an issue in video mode.

This does look like a pretty solid camera to me overall, so if you’re considering it, I’d say go for it, and would love to hear your detailed feedback (I’m sure the community and perhaps LabPano would be interested as well).

1 Like
12

The specifications looks very similar to the Ricoh Theta X that I have been using for the past year, and at less than even the repair/replacement cost (~$600), so this seems like a excellent “budget” option in comparison. (Yes, Theta X bit slow to get the GPS started, (so sometime no GPS-data in the first or second clip).

With the Theta X the USB-C connection if further away from the lens, but a standard cable end up more visible in the pictures than wanted; and found a my self in need of a cable with angled connector, e.g for when mapping more for longer than 30~40 min. (In my experience the Theta X is more stable when swapping batteries, and it a bit more convenient to do so then managing a cable to a pocked power bank…)
So from you experience, how much impact on images the does the a USB-C power cable have ? Have you tried it with a angled connector ?

Also as the leases are typically very sensitive, (yes I damaged my camera lens by letting it slide around it in a cardboard box for the drive home); do you know if LabPano offer any Lens Protectors/Guards as an accessory ?

13

Yes, the package comes with a soft cloth lens cap.

1 Like
14

On the importance of good quality pictures =
Yes, on the right that is a defibrillator.

15

Absolutely, and not only that! House numbers and street name signs are key features that must be legible in the imagery, otherwise this whole 360° street view capturing exercise is a pointless undertaking. This is why the minimum resolution for 360° imagery is 32 MP, which at first glance 8K seems to deliver. But again, we are talking here about bare minimum resolution only. Sensor noise, exposure time, and digital artifacts have to be acceptable too. So, before I spend this kind of money, perhaps even more, for basically a single purpose camera, I got to be sure that it covers all requirements well enough.

@boris Having said all of the above and since you have been provided a sample camera from LabPano, I hope that maybe you can be more successful in forwarding our recommendations to LabPano to make the PANOX V2 really worthwhile (maybe even a champion) for Mapillary. Because the hardware seems to be up to the task but the software needs to address Mapillary’s use case better. And, I am pretty sure that it is doable to get the software right. It just needs some more work. Video is really just a workaround for Mapillary, nothing more, nothing less.

3 Likes
16

Hello!

I’ve received the Panox V2 yesterday and have started trialling it with Mapillary. It’s my first 360 camera, and I’m keen to use it with Mapillary to help in my efforts on Openstreetmap.

First impressions, big gap in detailed documentation and community support.

Just on the consumer side and recording fun 360 videos, I figured out that videos must be ‘stitched’ before they’re useful. This can be done in device. I don’t know the history of 360 cameras, so I assume this is typically done on a computer. That’s all fine. It’s a bit slow to stitch videos on the device, slower than real time, but it’s fine.

I’m using street view video mode for my Mapillary captures. This is my main use case. 8k 5fps. Mounting the camera on top of my car. Captures aren’t coming out great. Can’t read street signs 5 meters way, for example. I don’t understand what options like ‘Bitrate Multiplier’ or ‘Video Quality’ does. I’m a video professional. I like to know bitrates and encoding settings, but they seem to be abstracted in the camera menus. The videos seem to be coming out already stitched and ready to use, but I don’t know if I’m missing a processing step to make it higher quality. Also I think the horizon isn’t locked like in normal video 360 mode. Without documentation, it’s hard to tell what these settings actually do or what I should be doing to get best results. I may have to do trial and error. Maybe my expectations are too high for a palm-sized device, but I’d like to read street signs and house numbers. Advice would be appreciated.

GPS seems fine. Google Street View Studio does complain, but I can’t see a problem with the actual data. It’s accurate within 3-5 meters and is very quick to acquire a lock for me. Under 5 seconds. I will however be using an external GNSS device, and build it into an ‘all in one’ mapping tool, so that I can do accurate surveys as well as 360 captures.

My tests are still processing as of this post, but here’s a link: Mapillary

I’ll continue tests over the next couple weeks.

3 Likes
17

8K is way too little for any meaningful data collection, like you have already mentioned for things like street name signs and house numbers. Keep in mind that 8K ≈ 33MP may seem like a lot of rays (pixels) projected on a plane but it is very little on a sphere. Resolution requirements increase cubed for spheres. Please try capturing in image mode using time lapse.

18

quality looks quite good to me…

Resolution just not good enough to read house numbers at 20 metres distance. You probably should not compare these devices with google streetview equipment. Lets wait a few years for 16k

19

That’s fair. I’ve started having good success with it in real runs. The quality actually is absolutely fine. Sure it’s not Google, but it does 90% of the job, and it’s invaluable to be able to capture up to date imagery of areas that have been missed by Google for years.

I’ve had a couple glitches, so some tips for anyone else looking to use the Panox V2:

  • I’ve had a capture file fail processing for both Google Street View Studio and Mapillary. Not sure why yet, but I had the file segments set to 30 minutes, so it’s affected the whole 30 minute file. I’ll be using 5 minute segments for now on to reduce the impact.
  • The Street Video recording mode doesn’t keep the horizon level, so ensure it’s mounted straight.
  • Phone connection to the camera can be difficult. I’ve found success by manually enabling the hotspot feature on the camera, connecting to the hotspot manually on the phone, and then opening the app.
  • If the image on the phone preview freezes but hasn’t disconnected, the recording has likely frozen too.

Despite the above, I’m very happy with the camera. GPS accuracy seems very good. Not having to process captures is fantastic.

1 Like
20

Thank you for sharing your suggestions and success @JamesPain - that’s great to hear!