Thanks @kaylesk, I have played around with the dng-file in Raw Therapee, which is an open source image editor. Xiaomi have definately done a few tricks to turn the sensor data into what we see. But with my artistic differences I came pretty close to the original result:
The Raw Therape file which contains my changes.
simonmikkelsen.dk/files/IMG_20171116_122545-rt-2.jpg.out.pp3
The rendering of the result.
simonmikkelsen.dk/files/IMG_20171116_122545-rt-2.jpg
I think it is possible to get a better result than the camera provides. If a picture is a bit over or under exposed it is also possible to get more details into the final result, but of cause not as much as from a real DSLR.
Raw Therapee supports copying the settings from one image and pasting it onto many. Also batch rendering in multiple threads. The images can then be exported with low compression and stitched as described in another place in this forum.
Camera control via command line/python (proof of concept)
Prototype for 2 seconds shooting with external intervalometer. Only few images 
Very low battery plus outside temperatue around 3 degree celsius and only one USB cable 
Small board with PIC processor and relay.
After turning on PIC switch relay (relay is connected to two bottom pins of mijia), wait ±2 second and repeat.
Sometimes saving image takes nearly 2 second and camera “skips” one take. Will try other card (now SanDisk Extreme 64GB) and/or fixed shutter speed/iso.
This looks awesome. I am looking forward to see future results of your tests!
I’ve got good news about the 2 second interval of this Misphere.
In the Misphere Facebook group I expressed my disappointment about the fact that 5 successive firmware updates did not fix the 2 second interval issue. I was then contacted by Harry Wang of Xiaomi asking if I wanted to beta test a firmware version. In the first beta version the 2 second setting had a 2.3 second interval in practice. The second beta version had a 2.1 second interval average. I tested this with the wifi on and the smartphone screen off . As soon as I put the screen back on I could hear (beeps) that the interval speeds drops a little.
In the facebook group Xiaomi also responded that GPS location is only set once in 10 seconds. That’s why I saw that location of images was not always correct. Xiaomi promised to look into this as well but this issue is not yet fixed in the beta firmware.
If anybody likes to test this version (at their own risk) the firmware can be downloaded here. If you want to know how to update firmware check this site and read under “how to update by SD card” . In this beta version you have to select the 60s intervalometer setting which is in fact the new 2 second interval. In the alfa version this will be fixed of course.
Cheers
Peter
Comparison of Xiaomi Mi Sphere (in camera stitching, full res) to Ricoh Theta S from a Mapillary workshop yesterday in freezing Helsinki. I used the Mi, walking all the time with handheld monopod, Theta user stopped and had a pole from the ground for better stability which was necessary due to HDR mode.
Location 1:
Mi sphere
Ricoh Theta
Location 2
Mi
Theta:
I’m using it some time now, with a 90° connector, and it’s nearly invisible. The only thing one has to take care of ist the stupid latch which will be visible when not attached firmly to the cable.
see my recent uploads last week
One problem persists: the mediocre stitching which I solved by buying the superior stitching software called “MiSphereConverter” for Android (totally worth the price!)
Another thing is that the automatic horizon doesn’t work with low temperatures. I had to strip the afully wrong exif data and adjust the images manually (which is quite easy with said app)
Besides that I’m totally happy with it!
@mc64 That is indeed nicely stitched. Seems better then the PC app that I use. I just wonder if the android app can stitch thousands of images in 1 go. The PC app also crashes some 3-4 times stitching thousands of images.
The first results with the true 2 second interval seems promising. The camera did not crash and kept taking photos. I also just learned from Harry Wang that Xiaomi is working on the GPS issue. They are going to add a switch bar to app, which can select GPS communication interval. I do not know which GPS interval they are thinking about but I assume we can choose intervals that are far less then the current 10 seconds. Hopefully also something like “continuous” or at 1 or 2 seconds.
@peewee32 you can indeed. If there’s no need to adjust anything you can stitch all images even during import!
@peewee32 hooray, will try firmware. But first I must assemble camera back. Because one small drop equals big lens scratch
https://photos.google.com/share/AF1QipPwc8nvlVznPvDUJiBpO-_OdEKMNldybXHg_ez-d-1QFwy4M3sNvUDfE5PL4s96ZQ/photo/AF1QipPWlxP7vwYrCH49r8IHYF4T-yBsrAi2WNMeBI6x?key=MGVwS3hpMHpwUDduZzVMSnFHMHdFQWEwX0phX1ZB
Waiting for second camera
If you communicate with Harry Wang, please could you ask if its possible to add subsecond time to exif?
With full second time in EXIF and connection with external GPX, gaps between images are not equal.
Some details about my external trigger.
Switched from sandisk extreme to Samsung evo card. Don’t know why, but write tests on sandisk card showed after while “down peaks” of write speed. (under 7M/s) Samsung write speed was line (more or less).
And after tweaking waiting time of PIC closer to 2 seconds, it works OK.
Mapillary last sequence with first camera with ISO 50, shutter auto.
From my experience with the gopro it’s not good to have units under one second. I once setup the gopro to capture at 0.5seconds,when geotagging on GeoSetter it ended up placing two photos each 3 into the same location.
A deal alert on the xiaomi at gearbest is currently at 218 euros shipped from their Polish warehouse w-g-4, so no import taxes for those in Europe.
Yes, this happens (sometimes) even with 1 second.
But I don’t need 0,5 seconds capture. 2 seconds is OK.
My “problem” is this.
Start capture each 2 seconds and time is (example)
14:45:04, 14:45:06, 14:45:09, 14:45:11, 14:45:13, 14:45:16, 14:45:18…
As you see not always 2 seconds, because delay is slightly more.
I know, it’s just a detail. If they can not or do not want to modify firmware I’m fine with that.
If 2 seconds will work direct on camera with GPS from phone, I will be happy as little panda (courtesy of Bjørn Nyland
I also had simmilar issues with Sandisk cards, especially wafter they were worn some.
In my contact with Sandisk they referred to their speed test, but it would only write 1 GB (on a 64 GB card) and they would not accept a test that wrote more. I did not want to spend more time on it, except that I concluded that their speed statements are false, when they do not test the whole card.
For the same reason I switched to Samsung too.
my first sequence with it. Loving it! such a nice piece of gear. post processed on MjiaCamera on PC (crashed once in a while, but very simple with gyro calibration).
I’ve never been able to get gps working with the mi, but I tend to either be gps tracking or taking Mapillary sequences with my phone anyway. Then I interpolate the location from those for the 360° photos.
Worked fine on my side on the first time. Only downside I noticed is that when the intervalometer is short, 3 or more photos have the same coordinates. Will do more testing and see if it was due to the location (steep hill on the left)
Bought this camera too, for replacement of bricked Ricoh Theta S.
Moscow will be 360!
Wanted to get one too, but it had a black hole on one of the two cameras. Today I’ve got a new one - same. The first one was a Chinese model (telling from the packaging and manuals) and the 2nd one seems international (English packaging/manual), from Ingram Micro.
Had you guys have experienced this too? I’ve tried updating, but it didn’t help except that it enables interval shooting. Also, the images are fine when reducing the resolution, however the preview stays broken. Videos are broken no matter which resolution. Should I try with a third one or has someone an alternative for ~300€?