Giving up... for now

Well, I think I have given this program a decent run but for now I think it’s back to Agisoft for me. Have been using (or attempting to use) during the free period and also have paid for a couple of 3 month promo runs. It has been really hit and miss for me as far as successful outputs go- which is not acceptable for business use. My main gripes are as follows:

  • Lack of documented workflows for basic tasks. Example- see here:
    Faster workflow for adding ground control points
    This was asked for months ago but still not available. Step by step workflow is available for other programs- seems like with RC you have to fumble your way through all the settings, lots of trial and error and you may or may not get the output you want- not exactly what i would expect from a paid software. Someone needs to spend some time preparing a few PDFs of steps for basic processes with a simple explanation of what different settings do along the way eg. getting a DTM and orthophoto using GCPs as is available for other programs. There seem to be so many settings for all sorts of things… but what do they actually do? and what are the optimal settings for particular outputs? Unless you read all the forum posts you would have no idea- needs to be a lot more Help built into the program describing all these settings.

  • Crashes/Errors… what can I say?- with this program it is really hit and miss as to whether you are going to get a result or not. Program can run for hours processing a job and you come back and all you see is a big panel saying “unknown error” Well, that’s not very useful if you are trying to produce something in a business environment and are on a tight timeline.

  • Inconsistent outputs. Try to follow same workflow from one project to the next but the output can vary in a seemingly random fashion. What worked for last project fails on current project. Very frustrating to say the least. Have had this happen multiple times when trying to output orthophotos- sometimes they are properly georeferenced and sometimes not?? Why?? I don’t know- I used same workflow as I did on the last project.

  • Failures due to hardware? I can understand you are probably pushing computer hardware to it’s absolute limit to get the impressive processing speeds we are seeing but could you possibly put in some fail-safe mechanisms to stop the crashes? maybe turn down the power a bit!! I would rather have one successful output which took longer to process than multiple crashes trying to process projects even at normal detail- I don’t think I should have to run my computer in a refrigerator or freezer to get a successful output out of a program!?

  • This one may be a bit contentious but … ahhh what the heck… there seems to be a bit of arrogance on behalf of the developers as to the capabilities and stability of the software in it’s current state. I get this impression from many posts in the forum- yes I think I have read every single post… many posts where users are posting issues but it is put back on the user that it is their fault for software not giving optimal results- their hardware setup, their input, their workflow etc. A good example from memory was about poor output resolution of images compared to same output in other software-a forum user compared output of same input through RC and Agisoft- RC giving lower quality output using same images but the response is that “your images aren’t good enough quality” - hang on… he is using the SAME images in both software and getting a better result in Agisoft- surely that would mean it’s not his image quality that is the issue but something inherent in the way that RC is processing the images is the issue?? I have beta tested software and hardware on many levels for many years and I think that developers need to be very open to the fact that that their product may in fact not be perfect. Trust me… beta testers often are putting the product through a much more vigorous testing environment than can be simulated in a developers office- you need to be appreciative of this and use this feedback to improve your product, NOT blame the user for the issue (especially if the user is actually paying to use your product!).

I still believe RC will get to a stage where it is a competitor with other big name photogrammetry applications. From what I have seen from a few of my successful outputs RC is capable of generating outputs that are superior to other products, more accurate models (I’m a surveyor- this is very important for me!), in a much quicker time frame.
Maybe it’s just that there are a still a few issues with the stuff that i am trying to do- ie. modelling and orthophoto generation from drone footage. It may well be that the close range stuff works quite well eg. turntable projects, body scanning etc.

I just think there are a lot of issues which should have been addressed (eg. documentation) in the alpha/beta phase before moving to a paid product. My 2 cents worth. I might try again in 12 months or so.

Dear ,
thank you for your post. We value your opinion very much.

First of all. We are very thankful for all contributions to the forum of users and of our support guys. Our goal is not be arrogant in this forum. Our goal is to help our users to solve their issues. We definitely do not think that our software is perfect. Therefore please accept our apologies if there was written something arrogant in this forum.

We will try to find threads you are talking about and we will discuss them with responsible persons.

Let me just note that we have officially released PUBLIC selling of our product in February this year. Our software is still in Beta version and our competitors are on the market much longer time than we are. We are young team, and we are young company. We are doing our best to solve all issues and deliver all tools our users need.

“-Failures due to hardware?”: I must say that I disagree with you in this point. Of course our software tries to use all possible computational resources. However it is not the reason why it is so fast. The reason are different algorithms. Our algorithms work almost always in linear time to the number of inputs. That is what makes us unique.

“- Crashes/Errors” : yes you are right there are issues, bugs, etc. just like in any other software. Are you working with Maya or 3D’s Max? Well, I was using them some time during fbx support implementation recently and I got many HARD crashes. Of course, it can’t be an excuse for our crashes. We want to have our product as much stable as possible. All screenshots you sent are either fixed now or there are known workarounds / reasons. Moreover, we have many satisfied customers that are using RC on daily basis so they must have stable workflows.

“- Inconsistent outputs.”: it is very hard to react to this point because it is like when you come to a doctor and tell him that “I’m Ill, please give me a medicine”. He will have to do many tests and he will have to give you many questions first. We have to do the same first. If you wish then we will be more than happy to investigate with you these problems.

"Lack of documented workflows for basic tasks. ", " modelling and orthophoto generation from drone footage " : I completely agree with you in these two points. We must improve documentation, especially we must create new tutorials. We must improve workflows and tools for “modelling and orthophoto generation from drone footage”. Especially GCP placements etc.

The biggest obstacle in this direction is that it is very hard to find a reliable expert person that can help us to create high quality tutorials etc. Therefore, maybe, here is also right place to ask readers of this thread: If you want to help us with tutorials then we will be more than happy to support you with free licenses and maybe start a collaboration, just contact us.

Thank you again for the post.

I just wanted to mention that after a few auto software updates this past week (don’t know which days you guys released them) I have found builds to be more stable AND my once ‘wimpy’ hardware can complete a normal quality build with the standard amount of ‘image downscale’.

@michalj, you mentioned some of 's crashes have been fixed. Were those fixed recently in this auto update I did?

michalj,

Thank you for taking the time to respond to my feedback/ criticisms. I guess I write out of frustration- especially of late as I just can’t seem to get many successful outputs (maybe 1 out of 5 don’t fail on me) - this used to be better but I don’t seem to be having much luck at all in the last couple of weeks. Thus am very interested with 's post above… have there been any (?silent) updates in last few days that I may have missed? Maybe it’s a case of the latest build fixing things for some users and breaking for others eg. me :frowning: ?? It’s difficult to isolate to a particular machine and hardware also due to the licencing requirements- can’t just quickly change installation to another machine if main one fails processing.

I can appreciate what you are saying as far as beta software and bug testing etc. I have beta tested software (and hardware) in the surveying industry for many years and I fully understand where RC sits vs other similar applications. Probably the term arrogance is a bit too harsh- just get a hint of deflecting user’s issues back on themselves which I don’t think is helpful towards software development and bug fixing. In my experience a large group of beta testers using software in real world scenarios is far more likely to find all those odd little bugs and generally do a better job of breaking the software than the programmers can do. I know in my case there’s a high likelihood the results I’m getting aren’t ideal as I haven’t tweaked the software settings to suit my imagery and desired outcomes- but this comes back to lack of documentation which is acknowledged to be an issue at present. I would be quite open to developing a pdf workflow for what I do- ie. DSM/DTM and orthophoto generation using GCPs (construction surveying use) but I have lost inclination to do this lately as RC keeps failing to process anything of reasonable quality!

Am planning on building a higher spec workstation in the next few weeks so I will try again once I’ve got this up and running and see if that solves some of my problems.

thanks
pete

Hey

I clicked on the question mark on the top right of the UI, next to the ‘RC’, and it says I am running version 1.0.2.2158. How about you?

,
I’m on the same build.

pete

Hmm, that is a bummer… Before these updates my graphics driver would crash when doing normal quality on some projects and most definitely crash on high quality. My last 5 builds have run without crashes on normal and high quality.

Yeah, just tried another run and it failed on normal detail- it worked on preview quality however (though the output orthophoto is not useful).
What hardware are you running? I am using an Alienware core i7-2760QM @ 2.4GHz, 32Gb ram, 2x Nvidia GTX460M, SSD + HDD. I realize running a laptop is not ideal as cooling is not as good as a desktop but I run it on a cooling pad with 3 fans whenever I run these projects and it has worked “reasonably” well previously. I might try a clean install of windows 10 as my licence needs re-activating in the next couple of days (currently running Win7).

Not sure what is going on- is it something in the software that has changed? or is my data not adequate? Last 2 projects I’ve attempted have been ~400 photos from drone flights at ~120m AGL. Flying a bit higher than I usually do and not as much overlap as I am just trying to achieve a basic orthophoto- not super high resolution DEM etc. Maybe the lower quality (compared to other projects) of these images is causing RC to fail (photoscan still processes OK however- just a lot slower). The reported errors for my GCPs (have 8 GCPs spread over project area) are all quite OK as far as I’m concerned… mind you the project fails whether I’m using GCPs or not so I doubt this is an issue.

400 seems to be pretty low number to be having issues with.

I’m getting issues too, but last couple I’ve done around 700 its went out no problem on high.

I’ve got issue’s in the 2000 - 5000+ range while trying to get high detail scenes out.

some of my issues have been camera alignment, if you have bad alignment it needs a lot more ram.

but I’m also running 128gb of ram.

Another flight today- flew lower (100m AGL) increased sidelap to 50% and forward lap to 60%, same area but this time ~700 photos.
Aligned images Ok, reconstruction failed a couple of minutes in on Normal detail.

Seems I can’t get a useful output out of this at present (other than preview quality- which is pretty much just a lot of blobs )

are you shooting straight down?

I’m find shooting at down 45 degree’s with phantom 3 doing an orbits works pretty well. rather than shooting in the classic grid pattern.

I did one today that worked well, only around 250 photos though, got a high detail model in a hour or so.

now I’m just trying to add another 10000 ground level photos. and then I’ll have another 2000 of so heli shots from previous shoot to add after that. I’ll see how i go with all of these, but I’ll probably run into issues.

are you also flying everything at the same height?

I would suggest you want at least a few different heights. though i think more different random heights is better.

it may be that a combination of grid flying, and then few different obits at different heights will work well.

I’m still experimenting with this. and i should do some more testing.

but i think at the taking for photo stage, think about creating 3d model, and not an ortho map or mosaic. they just happen to be easy outputs once you have the model.

Thanks Chris,

I’ll give that a try. What are you using to do your angled shots or are you flying manually? i have been using dronedeploy to fly areas and it works pretty well, but only allows for nadir shots… maybe I could use litchi app… it allows for time lapse shots at whatever angle you want the gimbal I think.

I’ll do a quick test combining my 120m flight with the 100m flight for starters.

Also going to try downgrading to previous download of RC- if that can be done using same licence??

I seem to recall reading on another forum some guys getting good results from adding angled shots (doing stockpile surveys or construction stuff)- a bit of a no-no for old-school aerial photogrammetry- but maybe the software is so smart these days it not only handles this but benefits from it??

I haven’t been personally flying our drone so much. but its basically been put into orbit mode. and having timelapse function going. and then a number of manual shots as well. (the ones from the same spot aren’t ideal)

I do need to look into to some flight control software and try some grid flights once i get a chance. let me know if some recommendations. I’m thinking about running a couple of offset grids, face the camera 45 degree down. and then maybe running it a couple of times offset and facing different directions.

I also do heli photography as well. which is all angled.

I do think it has be thought about differently to the old way of doing it though. since we aren’t trying to make just a nice aerial photo that stitches well. its making a true 3d model. that just happens to be easy to take a top down view once its done.

I almost think about it as if you were taking the photos for a small object, then enlarge that to do a big scene.

are you flying a quadcopter or a plane?

if you have access to the site again, you could try a few orbits at 40m, 70, 150m (if its a large site maybe you need a few at each height that overlap). etc… just to help stitch all your existing photos together.

also pay attention to how much ram is being used. if its getting a error for out of memory. easy fix is to have a pagefile large enough.

Site is about 1.0km x 2.5km and I will be there for next 18 months. So plenty of time for tweaking and refining of processes (if only I can get something to work).
For grid flights I have used pix4d capture app (but it only allows rectangular areas and crashes a lot). Dronedeploy app works much better- allows variable shapes (can upload shapefiles for areas to map). Both allow you to enter flying altitude and overlaps and the software then generates flight path and flies in autopilot capturing images as required. Straight down images only however.

Combining of 120m and 100m flights didn’t work- I get this cuda error- 30- unknown error- so it’s not a memory issue by the sounds of it- a graphics card problem.

Grrrrrrr… another gripe- cannot wind back software updates. Will not allow uninstall and re-install of older builds- . So effectively I am left with a software that I have paid for but is no longer functional.

wrote:

What hardware are you running? I am using an Alienware core i7-2760QM @ 2.4GHz, 32Gb ram, 2x Nvidia GTX460M, SSD + HDD. I realize running a laptop is not ideal as cooling is not as good as a desktop but I run it on a cooling pad with 3 fans whenever I run these projects and it has worked “reasonably” well previously.

My machine is a dog compared to that! i7-2820 @ 2.3 Ghz, 24gb ram, Nvidia Quadro 1000M, pulling off my external drive from a USB 3.0 port. :oops:

i7-2820 @ 2.3 Ghz, 24gb ram, Nvidia Quadro 1000M

That’s very interesting… what OS are you running?
I will have to wait another week to change from windows 7 to 10 as I had to reactivate when I tried to downgrade the builds. Not that it worked anyway :frowning:

Hi

The 4xx series was really power hungry GPU series… that is why even with additional cooling its a problem to get them working properly…

If you would fill out the SIGNATURE could say it sooner :frowning: ucp.php?i=profile&mode=signature

one more option as one of last resorts, look for some local EXPERT that can CLEAN the inner part of your notebook as over the years there could be a LOT of dust on the fans inside of the notebook…