It will be interesting to see if this camera has any features suitable for use in Mapillary Ricoh's new Theta X 360º camera features 2.55" touchscreen, interchangeable battery and Micro SD card slot: Digital Photography Review
The specs read promising. Especially the 60 MP still image resolution sounds absolutely compelling because imho 32 MP (8,192×4,096) is the bare minimum resolution to consider when doing data photography/mapping. Anything below that is just useless and should be considered a toy or tourist gadget. And then, if the integrated GPS module is also going to be worth while then the Theta X is definitely going to be a must buy for any serious Mapillarian. 
The MSRP—if they can keep it down to that for the end customer in these volatile times—also sounds extremely competitive. 
Furthermore, Ricoh has learned a lot of lessons on software with the Theta Z1. I really hope that they are going to make all of their software experience shine on the new Theta X. 
2 Likes
Battery life Still image: Approx. 220 photos *6
That’s not much. But:
USB-C port on the side of the camera for keeping the device charged via a USB power bank without the need to swap batteries.
Great!
I wonder what the rate of the interval mode is. If it goes quicker than 1 pic every 2s. Would be great for bicycle applications.
Two image sizes are available to match the scene: 11K for a high-resolution bright-light indoor photography, and 5.5K to capture images easily and efficiently in other scenarios.
oddly specific information. So I guess they are using pixel-binning. This could be bad news for outdoor mapillary usage 
Also, their use of Android sounds good.
I hope they do not take “still camera” too serious and are able to limit exposure time while in auto exposure mode. Or else, photography while moving would just smear everything.
Android is notoriously unsuitable for this because of limitations of their camera2 API.
1 Like
I wonder what the rate of the interval mode is. If it goes quicker than 1 pic every 2s. Would be great for bicycle applications.
Yeah, imho 1 fps is the bare minimum for still image mode when moving. Anything below that is especially useless on cars.
Two image sizes are available to match the scene: 11K for a high-resolution bright-light indoor photography, and 5.5K to capture images easily and efficiently in other scenarios.
oddly specific information. So I guess they are using pixel-binning. This could be bad news for outdoor mapillary usage
Hmm, indeed, they might be pixel binning at low light conditions in order to reduce noise/decrease the ISO level, or to shorten exposure time. Well, this is at least how I read it. However, this may also be a hint at temporal interleaving when shooting video or to get up to like 2 fps in still image mode.
Anyhow, I am hooked and I am seriously going to consider to buy the Theta X in 2022.
Bad news: the minimum interval seems to be 6 seconds: Ricoh Theta X Review: NO STITCHING NEEDED for 5.7K 360 videos + 11k photos - 360 Rumors
1 Like
Re: 6 seconds: I was afraid of that. So video and time lapse (if applicable) would be equivalent to previous 360 video. The higher pixel count and presumably processing power would only be useful for still spherical photos.
Bad news: the minimum interval seems to be 6 seconds
If that is true then it is a bummer. 
Well, I guess that 6 seconds is okay for the markets they are trying to target with their marketing campaign, like real estate agents and hotels. But, it is definitely going to be a deal breaker for me as a Mapillarian.
However, I am pretty sure that this could be remedied with appropriate firmware. I can hardly imagine that hardware could be the bottleneck here; neither in terms of encoding speed, thermal budget, nor write throughput on modern SD cards. The reason why this relatively long 6 second delay may exist is presumably because stitching is done on the camera’s under powered CPU. As a Mapillarian, I do not care if I get two separate images per second (assuming they still use two sensors) that I would have to stitch together perhaps offline, or maybe on an external machine. Yes, it would be nice to have a complete 360° image ready to upload from the camera but for Mapillary’s use case 6 seconds are way beyond anything useful (unless you really want to snail around streets; good luck with that). Seriously determined people can automate offline stitching but they cannot automate 1 fps into the camera. So, I can only hope that Ricoh is going to acknowledge and recognize that a) there is a market segment that needs 1 fps or more in still image mode and b) that there are ways to deliver that, even at the expense of some inconvenience.
1 Like
I found on-camera stitching very convenient, as one does not have to deal with software then.
Open source software in this area tends to not be so optimized and requires more manual labor.
Official sotware from both gopro and insta360 was always horrible. Heck, you cannot even download gopro’s stitching solution without that stupid Microsoft market account.
And you need that specific NVIDIA GPU which in turn have crappy closed source linux drivers.
Even worse, the mobile apps. The one for the Insta360 is full of US and Chinese spyware:
https://reports.exodus-privacy.eu.org/de/reports/com.arashivision.insta360akiko/latest/
Someone tested it and found 3s for 11K pics. Some improvement, but still too slow:
It seems possible to get 10fps 8K h.264 Video as a plugin developer:
So maybe the limit is the JPEG encoder?
8K is about 30 megapixels.
11K is about 60 megapixels.
So purely from a pixel-rate point of view, 11K might be 5fps…
Might be there’s another limit in the h.264 encoder.
Nevertheless 8K10 sound pretty attractive. Hope the video codec doesn’t impair the quality too much.
Update: the quality is very bad, lots of visible block artifacts: RICOH THETA 8K Video Tested For Developers. It's Great. - YouTube
2 Likes
Thank you for tracking it 
Official sotware from both gopro and insta360 was always horrible. Heck, you cannot even download gopro’s stitching solution without that stupid Microsoft market account.
And you need that specific NVIDIA GPU which in turn have crappy closed source linux drivers.
I can feel your pain. It is as if companies deliberately want to hurt their customers instead of heading for broadest adoption through compatibility, interoperability, and accessibility. 
Unfortunately, too often too many marketing and bean counting wise guys have too much decision power in product development and management. Most of the time these people do not get the point of their job. They think their job is to save money, though instead their job is to make money by enabling their R&D department to put out the best product on the market there is and then price it accordingly in tiers. There are two types of companies in this world: Those who are in the business of saving money and those who are in the business of making money. Guess what company I want to work for and buy products from.
So maybe the limit is the JPEG encoder?
No, I do not think so. Basically all general purpose SoCs since 2015 have hardware accelerated JPEG encoders (and some combination of video encoders). Besides, JPEG encoding is computationally really cheap, even at 60 MP, compared to h.264 or anything like that. Hardware vendors have already mastered fast JPEG encoding almost 20 years ago. So, I am assuming that the biggest bottleneck is probably going to be the final equirectangular re-projection step. Since the two images that the camera(s) make are always going to have the same relative position to each other, there is not much need for true stitching here either. So, the stitching step could probably be greatly speed up by simply blending overlapping pixels. Re-projection could/should be implemented in GPU. Granted, I have not implemented equirectangular re-projection on a GPU but I would guess that any decent iGPU should be able to do that in less than a second at 60 MP. But hey, that’s just me speculating. In any event, it would be nice to see some info on that from Ricoh.
Maybe they are doing much more than just saving the pic, because, even for 5.7K images, they claim 1.2 seconds per picture, whereas 30fps for basically the same sized video.
1 Like
@MikeN, please add a tag “360° cameras”: 360° cameras - Mapillary Community Forum
1 Like
1 Like
Have used the 8K10 and 8K2 modes for a while now (after updating the firmware), and the resulting .mp4 files work fine with the mapiality_tools (2.10+), and most importantly the the overheating issues are almost non-existent when using these moves (at 25~30°C). 8K2 is fine for walking and the battery seems to usually last for over an hour, but while driving the camera needs to be set to 8K10.
A bit of a pity that there are no 11K2 mode. It is however odd that the App cannot download 8K pics/movies from the camera over WIFI. It is possible to fetch these using a USB-cable, but I found the process that inconvenient; much easier to just store everything on a large SD-cared (512GB) until one gets home, and can move the SD-card to a reader and use mapiality_tools directly process & upload the video.
The app can be used to configure the camera, when WiFi is on; but 8K have required the WiFI to be disabled, and this causes issue with the App. Using Bluetooth remote shutter button works fine though, but then you’d want to be able to hear or see the camera - and that can be problematic when the camera is mounted on a car…
Have had some issues with the camera suddenly dying when car mounted, and with battery and USB-charger connected; with only USB-charger connected things seems the camera seem more reliable. Regardless with any crash or sudden power loss the current clip is generally lost (0 bytes).
1 Like
Thank you for sharing your experience @brunofilmklov! Sounds promising.
Could you please share here some Theta X examples on Mapillary? Thank you!
Here is are example of Theta X in 8K2 mode (probably), helmet mounted or on a double invisible-sick with a 200g counterweight*:
Maybe not the best the camera can do, so more of what you can expect while mapping a area.
(*) Google’s street view studio really don’t like wobbly GPS data in the videos, and if the camera i stable in the X/Y direction the GPS precision seems to be reasonably good, probably around 3m or less on a good/clear day. On day I think yd’ä have to be happy with ~10m. (Same issue with the Cellphone’s GPS & Theta V, if I walk with it in my hand then I’d lucky to get precision of 7~10m; but press the phone to the shoulder and that would improve to around ~3m.) I’m using a RICOH Theta Stick Tm-2 + Insta360 1.2m selfie-stick for ONE R/ONE X/ONE/EVO + 200g counterweight; (yes, the Insta360 stick is clearly not intended for this - the upper part not robust enough we connected to the bottom of the other stick, but it does help with stability and have surprisingly lasted more hen 6 months).
BTW, found out the hard way that the lens is very sensitive to rubbing, as I accidentally left the camera on a cardboard box while driving home - this rubbed a way the anti-glare coating, which was sufficient to cause significant blurring. The only “repair” offered by Richo seems to be to replace the camera, at about 50% of the cost for a new camera, so guess could have been worse.
Car-mounted, 8K10 mode, with [PULUZ Lens Guard PC Protective Lens Cover Case for Ricoh Theta SC2 / S/V Spherical Cameras Accessories Case Sticky Lens Protector] worked allot better then I had expected:
Yes, the Lens Guard is barely holding on the the lens mount, but better than risking a cracked lens. Not recommended in rain, or dew, as the lens can protector collect water and/or get covered with water droplets - with more or less unusable pictures as a result. Though driving in a light drizzle might not preclude getting useful footage, (though it likely voids the warranty - especially if USB-charge is connected…)
I don’t think the improvement in resolution would warrant a upgrade from the Theta V, but the workflow improvement and ability to record more or less for hours with only 5~10 second stops every 5~20 minutes**, instead of spending around 1/3 of the time downloading clips to the phone’s SD-card makes a huge difference - especial when walking with someone else is.
(**) If also uploading the clip to street view, the clip need to be shorter - up to ~7 seem to work fine; this is not an issue with mapillary_tools 2.10 that will automatically split the clip during upload (batches if 500 images).
Have used the camera with a large power bank and a cable with angled USB-connector; and while this works, with the older 3.x firmware the camera used to hang about once per day and require a forced hard reboot - and that clip was of course lost. The 4.x firmware seems more reliable; but now there are issues with user interface getting stuck, though at lest the normal power down works. So after acquiring a number of cheep replacement batteries and a few a double-battery-charger, I find stopping and swapping batteries more convenient, then dealing with the powerbank and cable combo.
1 Like
Thank you for sharing.
The optics look promising, despite the coating situation. 
However, the low bit rate at 8K video creates too many artifacts thus rendering the video/imagery de facto useless for mapping. While 8K panos are on the verge of what could be considered sufficient for OSM mapping with consumer grade equipment, it is all ruined by low bit rates. 
Things look so blocky that you basically loose half of the resolution. If this is the best quality one can get from 8K resolution then you do not need 8K.
What about still image time lapse imagery? Could you share some of these examples, please. Thank you.
Yes, you are correct in that there is difference in the image quality, as the 11K single photos seems to require about 11MB, and the extracted 8K video frames around 0.5~1.2MB.
Further we need to consider that Mapiallity currently have a fixed 2K max resolution for the projected image, so for now AFAIK there is a upper limit for the useful the camera resolution.
No, time lapse imagery is not natively supported anymore, but might be possible through a 3rd party plugin. (I believe there is a 3+ second startup time for taking a single photo, so yes you could probably approximate a 11K0.2 mode that way; however I believe a plugin-API have a bit more leeway in the frame and bit rate selection… Yes, Theta V used to have function, but not X. )
The bit rates for 8K2 and 8K10 seems to be adequate to me, given the limitations of the the camera form factor etc. I agree that the 8K resolution is a bit of a limitation OSM mapping, is it severely limits the ability to zoom in to read signs and get numbers and names. For the next major step in quality I think around 24~32K (and 6+ 4K cameras) are needed, and this implies a significant increase in physical size, storage space and cost. (And if I recall correctly, there are drone cameras that hit Mapiallity’s single image max size already…)