Offline rendering?

Offline rendering?

Anonymous
Not applicable
4,903 Views
23 Replies
Message 1 of 24

Offline rendering?

Anonymous
Not applicable

I get this is a beta. I get that there are several things that still need worked out. I get that not all features are available yet. But when I've been kicked from the upload queue for A360 three times in a row in less than 12 hours, when I don't even want to upload to the cloud, is ridiculous. I saw on another post that the reason the forced cloud upload was implemented was by user request, but is there any good reason why there's no choice allowed?

 

Ok, rant is over. Basically, I have 2 requests.

 

1) does anyone know if/when the ability to use this without needing to upload to A360 will be available? If it's already there and I've overlooked it, how do I access it?

 

2) Is there any user generated "fix" or plugin I can use in the meantime?

Reply
Reply
0 Likes
4,904 Views
23 Replies
Replies (23)
Message 2 of 24

Anonymous
Not applicable
According to one of the officials here there will be no offline version, for now anyway.
Reply
Reply
0 Likes
Message 3 of 24

murali.p
Alumni
Alumni

>>2) Is there any user generated "fix" or plugin I can use in the meantime?

 

Can you elaborate on this? Plugin for what? You should not get errors *while* uploading & registring photos to the server. If that is what you experienced, then you can turn on debug logs and share that with me. Go to Memento install directory and look for memento.cfg and edit the line

 

<level>Info</level>

 

to 


<level>Debug</level>

 

Regards,

Murali

Reply
Reply
0 Likes
Message 4 of 24

Anonymous
Not applicable

@murali.p I have my photos already in A360. I also saw somewhere in another post that It'd be best at this point to use the draft settings, so that's what I've done. I see that my project has been queued for processing, but in a few hours I come back and there is red text below the project saying an error occurred. The only thing I can do is delete/cancel it and try again. 

 

Is it possible I don't have enough images? It's a simple object I'm trying to make, so I only took 10 photos. Do I need to do more? Would that cause an error message like that?

Reply
Reply
0 Likes
Message 5 of 24

murali.p
Alumni
Alumni

Can you enable the Debug mode in the config? That would help us diagnose. If the mesh generation fails, you should get an email saying so. 10 images should be enough if each image has good overlap with its neighbouring pair. Alternatively, share the images with me on A360 Drive to test. And I can share with you images for you to test that are sure to work. Let me know.

 

Regards,

Murali

Reply
Reply
0 Likes
Message 6 of 24

Anonymous
Not applicable

That seems utterly ridiculous that this would not have an "offline" mode! Why? I would like to use this "in the field" and at the moment mobile data in Australia is $180 per 12gb. so if I am uploading 100-200 high quality photos every time I want to use this that is an insanely expensive exercise. Is it really true that there is and won't be an "offline" mode? With mobile and workstation computing power the way it is these days I would struggle to see why not?

....

Reply
Reply
0 Likes
Message 7 of 24

TatjanaDzambazova
Autodesk
Autodesk

Hi,

 

Its OK to be demanding, and we love being pushed to do better and more. That is one of the main reasons to have this publically opened Beta ( meaning : work in progress sowftare"

But i am not sure i am a big fan of the word 'ridiculous', There are people behind this software who work very hard to make it happen.

There is a reason why we chose this type of processing, and you may agree with it or not.

So let me try to explain the thinking behind what we do.

 

If the photo to mesh and scan to mesh worked in offline mode, not reaching to the cloud but fully on the machine, given that it is a heavy computation task, to start with, you will need to have a very high spec expensive machine to run that. Then, you will need to wait for many many hours or if you are lucky, for days (depending on number of scans photos) in which time you will not be able to close your machine, bring it home or else. and you will be probably bearly be able to work at that time with other applications on your machine. 

 

This is where the promise of the cloud processing is:

be able to use the thousands of Amazon servers to crunch the heavy duty data processing, while not forcing you to wait, keep the machine open while its processing the data, nor forcing you to own a really expensive machine. With cloud processing, you will be able to do it from any standard spec machine, close it after the photos/scans are uploaded ( the speed of that only depends on the strength of the internet) and do other work or go home.

 

Now, two things are importnt here:

- as we are not in full production, we havent had dedicated enough servers for processing when we launched the Beta and we got such a high volume of jobs that we frankly didnt expect, that it brought us to a crunch with users waiting in a queue. We made improvements in the past 4-5 days and the wait times have bee be improved with every day. Additionally, there are some upcoming technology improvements that will make this really fast.

so, its one's choice how and where to process the data - we started by doing it on the cloud as that has many advantages (even if not fully visible with the Beta) and it allows us to offer strong tools with very flexible business models instead of old fashioned perpetual licenses.

 

- the requirement to do work on location is well understood as it has been demanded by professioanls who capture on remote sites where thereis no internet. For that they didnt necassarily need full high definition processing of the data on site, but just a proof that they have well taken the photos so that when they are back, they process the full res mesh on proper connection. For that a local stitch mode to show the user that they have captured the object well. We will be working on that. 

 

That is our rationel.

One of the pruposes of making Beta stages of sowftare IS to evaluate intial decisions, listen to requirements and see if we need to amend. We havent and dont say we will NOT have an offline mode, we will listen and evaluate how important is this to majority of the users, given all the downsides of the offline mode and th epromises of the cloud computing. but this is not a one button switch that enables to swicth to an offline mode. We had a choice to do it offline or online and we placed our bets on the future of cloud computing. 

 

So, thanks for the feedback, looking foward to more, but do understand that there is thought behind what we do. You just might not agree with it.

tanja

Reply
Reply
Message 8 of 24

Anonymous
Not applicable

HI Tanja,

 

I understand perfectly that there is a team of people behind this program that are doing what is quite frankly utterly AMAZING work. This program (and the idea behind it) is nothing short of mind blowing and brilliant! So don't take my criticism of one small aspect of it as a complete write-off of the whole program. I have spent quite a bit of time looking into this program to get as good an understanding as I possibly can on how it is currently working including watching over 3hrs of the YouTube tutorials and thoroughly understand the meaning behind Beta testing.

So let me elaborate a little on my initial comment of it being ridiculous that there is no “offline” mode.

 

Firstly, my reaction was to willohmsford’s comment of “According to one of the officials here there will be no offline version, for now anyway.” And I know it says “for now anyway” and that is a beta test. But notice that I asked multiple times was this going to be the case? And offered the explanations that even if it was possible to get an internet connection on a site it could be A) a VERY slow process given the slow mobile upload and download speeds when using mobile data. This is a screen shot of the mobile data test I just did and the site I am on right now have their own tower on site for their internet access, so the speed here is actually pretty good compared to an average land line connection in Australia.

Ping Speed.PNG

 

And B) an insanely expensive exercise given that when the product is officially released it will most likely cost a small fortune to use in a commercial sense, like all other Autodesk products, plus then we would be adding the expense of insanely priced mobile internet here in Australia, this could be a huge hindrance to smaller businesses in particular when considered over a number of years vs the expense of buying a decent computer to do the number crunching. And keep in mind I am not thinking in terms of using this once a week or once every couple of days. I am thinking in terms of using this on a daily basis and multiple times a day at that. So if it is going to cost the end user $XXX per month or $XXXX year then $x each time you want to process a job, these costs can add up fast to be far greater than a decent laptop or PC. Maybe I am thinking of this in a totally different light to the development team though. Maybe this was never intended to be used on such a regular basis. But given the examples in the tutorials of the Smithsonian and archeologists maybe I am not? Then add to that the cost of the other programs like AutoCAD, Inventor, Recap Pro, etc. you can possibly appreciate the cost implications of trying to use this.

 

Now also take into account the fact that most commercial users of this product, the users that would be likely to be using this in remote places away from good/cheap internet connections, would be most likely drafters, designers and others that already use massive data sets as part of their job, e.g. FEA, point clouds or massive 3D BIM and other models, the fact is they will already have a pretty beefy computer to run the numbers for their existing work. Also consider that nearly every office has at least one “spare” computer, the fact that you might need to set one computer to working on a job like this for a few hours or a day probably would most likely be of absolutely no consequence at all. Now I will be perfectly open here and say I may be massively underestimating the amount of computing power to do the number crunching on things like this, but I would assume it would have to be comparable to FEA and there are plenty of offices out there that do that in house.

 

All that said I can see the HUGE advantage of the cloud processing. I can understand the choice to make the main function or default function to be set to work in the cloud. But to not include the option for the user to be able to work offline in a program that they will most likely be paying a large sum of money for, or to use, seems totally ridiculous.  And this is a criticism that is not only leveled at this program but it is the way that a LOT of programs seem to be going these days, where they can’t be used without some form of internet connection.

so all that said, in the interest of trying to be a little more constructive in the development of the program, here is a couple of ideas for you and the development team to kick around.

 

  • Could you implement the OPTION to use the program offline? This would give the end user the option to run a job off line if needed or if they have a computer that can crunch the numbers. This would allow people to at least test their systems to see if they can do it. If not then they have the option to buy a better machine or only use the program online. and just keep this in mind before saying "do it when you get to a better internet connection", i do fly in fly out work on a mine site currently and spend 14 days straight on site and then 14 days off site. so if i were to do a job here on the first day where i wanted to use this process and i had to wait until i got to a decent internet connection i would have to wait 14 days to get home process the job. then wait another 14 days to get back to site to use the data. so maybe that option isn't so practical.

 

  • Given that servers are now pretty common place in offices, would it be possible to implement an option to process the data on a local server to be able to free up a workstation to allow a user to continue working while this is being processed?

 

  • Would it be possible to do the job in smaller chunks? For example, when you use 3D Laser Scanners to take point clouds on a job site, you need to take the scans from multiple locations and then stitch them together in a program like recap pro. Would it be possible to have this program do something similar? So the user might be able to load the job (photos) in smaller chunks and then stitch the smaller chunks to make the total job. Similar things can be done with panorama stitching software to make the process faster. This might be an answer to doing the task locally without needing the cloud.

 

Please try and understand that I am making these comments based on a view of using this software in a daily commercial process in remote areas like I am currently working in. and my particular circumstances are not that unusual in Australia. But I understand that in the grand scheme of Autodesk consumers I would most likely be a very small minority and possibly not even worth considering when it comes to the overall implementation of these things, so understand if the desire to work offline is not taken into consideration.

Reply
Reply
0 Likes
Message 9 of 24

Anonymous
Not applicable

Josh, have your tried VisualSFM  and Meshlab?

Reply
Reply
Message 10 of 24

Anonymous
Not applicable

I did a whole tutorial on that option.

 

Http://youtube.com//D6eqW6yk50k

 

Reply
Reply
0 Likes
Message 11 of 24

Anonymous
Not applicable

i have downloaded mesh lab, but haven't had a chance to play with it or learn to much about it yet. i will look into it more when i get the chance.

Reply
Reply
0 Likes
Message 12 of 24

Anonymous
Not applicable

No you didn't 😛 Smiley Very HappyCapture.PNG

Reply
Reply
0 Likes
Message 13 of 24

Anonymous
Not applicable

Sorry I had trouble pasting the URL on my phone so I tried to type it by hand.

 

https://www.youtube.com/watch?v=D6eqW6yk50k

Reply
Reply
0 Likes
Message 14 of 24

Anonymous
Not applicable

yeah figured it was something like that. thanks for the link. that looks like a good indepth tutorial. i will have a watch when i get time.

Reply
Reply
0 Likes
Message 15 of 24

kintaro_takanori
Participant
Participant

i give this a try, just like i did for 123d catch. i absolutely dont like the cloud based service. i dont like to share my privacy with anyone, online. this is very important to me. that being said, i will try memento because autodesk is just awesome and i use maya and other software from them. i think its great what they do but forcing people to upload photos (meta data) to a server that, which consumes time that nobody should invest, is just against the stream. i mean the time spent on uploading and processing huge amounts of data over the intenet then sending it back and forth is just too slow. the only reason why autodesk is doing this is most likely because collecting data &colon;) im very much against that.

 

greetings

 

ps: yes i want offline rendering, please!

be the best you can be...
Reply
Reply
0 Likes
Message 16 of 24

Anonymous
Not applicable

Hi Josh,

even if the chance to have an offline processing could seem attractive and less expansive, that would result in a kind of wish for which I would say " Choose wisely what you wish as you could get it!". I think only one of the user here in this thread raised a valid question "Have you ever used Visual SFM and Meshlab?"...well you replied "I have not, I downloaded it but I did not have time to play with it". Let me say something, I did it, I am still "playing" with them, I can assure you it is not a game to "play" with and working with them will make you better understand why you would not like a local machine processing. Briefly:

 

Let us start from Visual SFM point cloud extraction. This process itself is the first obstacle, there thousand algorithm you have to study and try before you can have even a rough satisfying results. You can find some of them online, but the most reliable ones are hidden in thousand of scientific paper you have to read deeply before you can understand how they actually works, and to read them you should first have access to this journal which most of the time are not open access but you must pay for them (so first out of pocket!). Then, when you will find a very good one, you should buy or build a system able to use that algorithm (second out of pocket) and if the algorithm is good as Memento's one you should leave your system computing for hours (assuming you have a very expansive super computer or so) or days (assuming you have a standard system), resulting in another out of pocket for electricity bills (in Italy you would need a loan only for this!). Once you have your sparse point cloud, you should have another algorithm for the dense point cloud generation that means having the same workflow as above for the sparse one. Now, after you have got your final results, you should process everything through Meshlab. Same thing as above, studying tons of papers to find a good point cloud to mesh algorithm, a good filter for noise data filtering, struggling with selection filters and so on. But let us imagine you use a simple Gaussian mesh generator to generate your geometry from point cloud. Even if you used an octree depth of 15 or so, the system you would need to process your data would cost you so much you have better rent it, and the time it would require to process the data (assuming you want a Memento-like result) will be so long that your nephew will probably see the outcome (just joking :P...but believe me it is a looooong time). Same here as above, electricity, a very high end system, cutting edge algorithm and papers/books to buy...
Finally, I can tell you something, I work for an University and we do have super computer but cloud servicing still results for us the most affordable e fastest way to have very good outcomes.

 

I hope this could be of some help.

 

Best,

 

V.R.

 

Reply
Reply
Message 17 of 24

Anonymous
Not applicable
I understand this on the side of processing power; however, if a company wanted to use this software in a commercial environment, the cloud system wouldn't be an option. Placing data of a sensitive nature for a high profile project on a cloud throws so many red flags up it isn't even safe to snicker at. Is there a way around this or an option in progress?
Reply
Reply
0 Likes
Message 18 of 24

Anonymous
Not applicable

1.2mbps is pretty good compared to Mexico. I get 0.31mbps maximum. Am running a MacPro (2013) 8 core and 20 GB Ram, 2 d300 Firepro video cards and multiple SSDs... but all that costly firepower is useless if I have to upload to a server. Will look elsewhere... Adios Autodesk.

 

 

Reply
Reply
0 Likes
Message 19 of 24

TatjanaDzambazova
Autodesk
Autodesk

Hi Jim,

wow , you are fast to say goodbye :). No problem if Memento is not for you, that is why Beta programs exist, to test and see if it would be somehting that fits your needs.

 

just to be sure this is understood:

- we only upload to the cloud to Create 3D, every other operation in Memento is local.

- We do that to eliminate the need for high end machines and as we work on distribution, to provide super fast reconstructions (once we are donewith the  distribution). Even with a strong machines, when you have to generate 3d from large data it will lock your mahcine for days and you will not be able to close the machine to go home or do anything else on it while it sweats over the data. 

- We DO have an offline version of Memento as well that we might offer when we launch ( This spring) but we are currently not convinced that it is a better user experience. we do have many conversations with users like you who want that, to understand well whats important to them ( for some people speed is important, for others, ability to be flexible and close machine while somehting is processing, etc etc). Informed by those conversations, we will decide how to offer the offline version and how/ what to charge for it. 

 

Finally and most importantly: creating 3D is just one little portion of Memento's value proposition. . Memento creates meshes from photos, laser scans (alpha opens next week)  and structured light, but thats the begining. Being able to super easy, pleasurably ( our mesh streaming engine can handle 2Billion polygon meshes) smartly clean up, fix, edit, sculpt, retopoligize, decimate with baking of textures (so that assetts look awesome on AR/VR that cant handle big meshes), prepare for 3d printing, create smart videos, make reports, measure, compare etc are tools that one has to go to many other sowftare to use, in order to create USEFUL and USABLE 3D meshes for downstream use.

mementos value is in providing all that in one application that will be, once we start charing, very affordable. 

 

 

thanks for trying and thanks for your feedback!

tanja

 

Reply
Reply
Message 20 of 24

Anonymous
Not applicable

I do like the appearance of memento but the offline version sounds amazing, I hope doesn't stay server based. I'll be waiting for this amazing app.

Reply
Reply