New Insta360 x4 with 8k video

@Hol_Ger I’ve just watched some of your latest urban imagery with the X4. The 72 MP resolution is really something! :+1: Makes you want to look for more.

And, I also like your backpack setup. :grin:

Thanks for the many facts. I always can learn from your posts.

Yes, I take the time for post-processing my photos, more or less. It depends on the beautiness :wink: of a location. And I want to turn into all directions in a scene without a tilting horizon. See my previous post New Insta360 x4 with 8k video - #12 by Hol_Ger. Quite the opposite of “how and when sensor buckets were excited by the physical interaction of photons” :wink:

These 72MPs are my first try. The images are really great, but require sharpening in a post-process. And this is really, really slow because of the image’s size. It takes about the same time as the hike :open_mouth: But this is mostly required with X4. At least its “pure shots” (a little bit more like RAW) are by far not so sharp as GoPro Max’s. And you need even more light as for 18MPs. So many of the images are motion blurred when I walked in forest especially when I walked faster. Also the sharpening impaired scriptures on signs. But from now on I will prefer 72MP whenver possible.

There are more impressive photos than this you linked to. See Mapillary or Mapillary. Landscape motives are much more demanding than street photos, because of all the fine structures you want to preserve.

If you ask for 2 fps in image mode, you get exactly 2 fps

This is of course very nice. But for hiking not necessarily required. I do not know why X4 is so much slower in this point.

you can also simply lock sensitivity and the firmware is going to be quite well up to the task balancing exposure time

What about the other way round? Locking exposure time? I mean setting the longest exposure time and then increasing sensitivity.

Unfortunately, you cannot lock exposure time with the latest firmware on the GoPro MAX in time lapse image mode. The only mode that enables you to freely set exposure time is the single picture (full manual) mode, which is basically useless for Mapillary unless you are willing to tap your phone or the camera continuously. The only sort of automated mode where you can lock exposure time is the video mode, and that in a limited way only. You can lock exposure time to match the full frame time. For example, you can configure to shoot a 25 fps video and you can lock the exposure time to 1/25 s on every frame. You can also let the firmware choose the exposure time automatically, of course no longer than the frame time. You would usually use the former to get motion blur effects like on real film and to capture more light. The later you would use to crisply capture fast moving objects or to get sharp frames when the camera is in motion.
Btw, with the longest exposure time set, I think you mean to decrease sensitivity. :wink: Well, in time lapse image mode the camera — in fact both sensors independently — automatically continuously adjust sensitivity and exposure time. However, you can lock to a certain sensitivity range between ISO 100 and ISO 3200, e.g. ISO 100 to ISO 400, or ISO 200 to ISO 200, etc. But, like I’ve already mentioned above, you cannot lock exposure time in this mode. However, you can lock EV compensation but this is relative, of course. And, you cannot lock both sensors to identical parameters either, which is another huge bummer because this would have been like a natural option to have, I guess, but apparently the GoPro engineers did not think so. :person_shrugging: The only mode where you can lock both sensors and exposure time is again the full manual single image mode.

What I mean is not to use 1/50s when I set the longest exposure time to 1/200s and instead switching from ISO 100 to ISO 400 automatically. Is it increasing or decreasing sensitivity? I am confused. Thought it is increasing (100 → 400).

Is it increasing or decreasing sensitivity? I am confused. Thought it is increasing (100 → 400).

You are correct. If you lock exposure time, especially to some very short value, then the only way to compensate for lack of brightness past the locked exposure time is to increase sensitivity. What I understood by “setting the longest exposure time” is that you meant to set exposure time to the camera’s maximum value and then let the firmware adjust brightness by modifying sensitivity. In this scenario the firmware would usually be decreasing sensitivity because of the very long exposure time, which I think is 30 s on the GoPro MAX. Anyway, you cannot limit the exposure time in time lapse mode either, like by setting an exposure time range, similar to the sensitivity range already mentioned.

Because I favor low noise and I usually just walk when capturing in daylight, there is plenty of light to ensure short enough exposure times to capture almost everything sharply, even at lowest sensitivity levels. So, I limit the sensitivity range to ISO 100 only. At dusk however, the sensors quickly get into trouble and you have to increase the range to like ISO 400 (beyond ISO 400 it becomes just a noise fest). Unfortunately, the firmware’s fixed exposure curve clearly favors short exposure times (because you know, it is supposed to be an action cam to capture lots of action! :disappointed:) over low sensitivity, so it increases sensitivity rather overly quickly. Standard GoPro HEROs let you tune your own exposure curve.

Ah, digital sharpening! My next gripe with everything with fixed aperture. The GoPro does sharpening in camera and you can set one of three levels, “low”, “medium”, and “high”, whatever these may mean (I usually go with “low”). The only plus side to this is that the GoPro is really fast at sharpening, it is basically instantaneous. Yeah, I really would not want to waste my time on sharpening in post‑processing. So yes, it has caught my attention that there must be heavy sharpening involved in these 70 MP images because these algorithms usually fail miserably on surfaces with lots of high contrast texture, like asphalt or roaring water (and then on top of that comes JPEG quantization, which also struggles with such surfaces). This is why I wish fixed aperture camera makers would finally start using smaller apertures but what do I know. :thinking: I think there is already a patent for years available to have a piezoelectrically controlled two step (could be more by now) aperture in this type of sensors. But, apparently there is some sort of licensing war going on why we have not seen these in consumer market devices yet. :person_shrugging:

X4 does only offer three levels in video, not photo mode.

In photo mode you can choose between two different processing types: INSV and Pure shot with INSV being much sharper than Pure Shot. With landscapes both does not meet my expectations. More urban motives look good. This is with in-camera sharpening: Mapillary.

BTW Here is my favourite GoPro Max sequence: Mapillary. Not so good as with X4, but not too bad. Would realy like to know whether it was post-processed in some way. But again the tilted horizon: Mapillary :disappointed: and too deep shadows: Mapillary. Compare the shadows with X4’s automatic settings: Mapillary. Even with backlight they show structures.

1 Like

WIll do some more tests today…

Could you make it running?

@Hol_Ger thanks for asking I feel like I miss something basic

  • small test today Timelapse and I dont understand how to Photo Export link

As far as I remember, GPS data gets lost when exporting as reframed video. With regard to X3 you can find instructions here: https://help.mapillary.com/hc/en-us/articles/11951588568604-Insta360-X3. Pay attention that the insv video file, not an mp4 should be uploaded. But I am not sure that it is working for X4 too because I read Insta changed the internal format between X3 and X4.

You can not export a video as photo sequence only Intervall photos:

1 Like

So, I started playing with sharpening in post‑processing on GoPro low sharpened images to see if I find something that I like. I am no sharpening algorithm expert, so I went with ImageMagick’s -adaptive-sharpen. It is supposed to be standard Gaussian sharpening. I do not know if it is good or bad. After reviewing some results, you could say that it makes features (or structures, if you like) more pronounced, especially in shadows. On the other hand, it increases noise on textured or bumpy surfaces, reduces the overall number of colors, and makes JPEG quantization artifacts more visible. Bad parameters can easily make things look unnatural. So, does it make GoPro images look better? Maybe? I do not know. Maybe it is just a matter of the right algorithm and parameters.

What do you use to sharpen your images in post‑processing? And, what parameters do you use? In other words, how do you select parameters?

1 Like

Like me. I fiddle around with -unsharp, e. g.

-unsharp 0.4x0.4+8+0.001 or 0.4x0.4+4+0.001

inspired by Efficient Image Resizing With ImageMagick — Smashing Magazine.

What I found so far: Light conditions have a big impact on the settings. Resolution too. Theoretically one have to apply different settings to images taken under different light conditions. That is really too time consuming when there are thousands of images.

Today I uploaded 72MP photos without extra sharpening. Some are more and some less satisfying.

I saw your new sequence Mapillary. One of your sharpening tests?

1 Like

This is something I have experienced too.

Thanks, I’ll check it out.

That’s right, my first shot at sharpening in post. I have no idea whether I made things better or worse. I just went by subjective looks of it. I used -adaptive-sharpen 16x3 :person_shrugging: Actually, it was already a bit too dark for capturing but I wanted to stretch my legs anyway and make best use of the time. To be honest, I have been expecting worse in these conditions from the GoPro. I have more imagery from this evening to come but it is too much for my current trickle connection to handle reliably. So, I may go with other post‑processing parameters or none next time. What do you think of it, the parameters and sequence?

After having acquired some experience with the GoPro MAX, in some situations, I am tempted to use my old compact camera because I am surprised and delighted to see how well the old camera performs compared to the MAX. In terms of quality, its sensor is very much comparable to the GoPro but I think its three lens optics and dynamic aperture give it an indisputable edge over the MAX (although I am aware that you cannot compare these cameras directly). The GoPro MAX’ quality reminds of phone cameras in the late 2000s. But, its time lapse mode enables me to cover more ground faster and excellent GPS simplifies my workflow. :+1: Plus, the GoPro has just spectacular battery :battery: life in image time lapse mode!

Ricoh Theta X has 8K Street View video capture mode, which makes it fit for Mapillary capture too.

Could you provide some references? I only know these one:

and they are of pretty low quality.

I am tempted to use my old compact camera

Of course. My DSLR makes much better photos than the X4 too. But when I look at them and try to shift the view …

the GoPro has just spectacular battery :battery: life in image time lapse mode!

How long does the battery last and with which settings?

BTW I deleted all photos uploaded yesterday with no extra sharpening and replaced with 0.8x0.8+4+0.001. Looks much better and the work was done from the computer overnight, not me. I wasted time with sleeping :wink: Mapillary

I have no idea whether I made things better or worse

Looks promising. But can be improved, I think. Perhaps trying stronger in-camera sharpening? See Mapillary.

@Hol_Ger
No, I haven’t been capturing with the Theta X much and no Mapillary at all. I bought it because I was completely fed up with Labpano/Panox.
The examples you give are made with the sun low at the horizon, that is a disaster for 360 cameras and against Google’s Street View guidelines.
Image quality has many aspects, which are you referring to?

Refer here for photos with Theta X, which are 11K, unlike 8K video for Street View / Mapillary.

More asthetic aspects :grinning:: good sharpness, wide density range, details in shadows, no desaturated look, no artifacts. As little post processing effort as possible :wink: Perhaps faster in interval photo mode. My X4 produces just one interval photo in 6 sec.

Interisting. What was it that you are fed up with?

Not always (open in browser):

I was one of the first buyers of the Labpano Pilot Era. It was an unfinished not properly tested product. Lens misalignment causing stitching issues. Software bugs and incompleteness. Their promise of ongoing development of the software was troublesome. Promises for functionality development were never realised. The Chinese ladies for international marketing and service were nice in our frequent contacts, but did not understand much of the technicalities, which of course created communication issues with the technicians. Worst was that their original development team appeared to be hired for the development project and left thereafter. Others have been tinkering with the software since, making image quality worse than it was before. Worst was that they tinkered with the auto white balance, which they never got in order anymore. Same with ISO, resulting in grainy footage.
Although their cameras are promoted as Google Street View cameras, they clearly have no idea of the practicalities of it. Plenty of places on earth of days with direct sun of 50C and over. Their and most cameras are not fit for that. There is no manufacturer that I know that tunes the capability of the GNSS hardware properly. After studying those issues, I have pointed them out to them but never got improved.
Their first Pilot One was a complete design disaster; plasticky and therefore breaking apart while on car mount. Separate GNSS module affixed on the bottom with the flat antenna on the side, a disaster for signal reception.

Not to mention their idea of auto virtual tour generation. They were doing nice SLAM developments, but that does not deliver awareness of walls. So, when they released it, it was not working. I do not think that they have pursued it further.

3 Likes