Messages

1

General / Thin Point Cloud - how big the reduction is, or any final default value result?

« on: July 13, 2023, 12:27:06 PM »

Good Morning administrators and everybody,

maybe someone would like to share information regarding editing point cloud results by filtering with the option of "Thin Point Cloud". From my perspective, the option facilitates us to reduce the number of tie points after the photo-alignment process as well. One of the purposes may relate to the matching strategy with our machine performance because excessive tie points require higher computing capabilities which sometimes would be problematic to the entire process.

The question is how big will the reduction be (or how many tie points will be kept and safely remain in the chunk) or what is the reduction procedure (25% or 50% etc.)?
 
For example, a project originally processed with a setting of tie point limit of 16,000 resulting in 36,000,000 tie points. When we execute the "Thin Point Cloud" filtering option, how many tie points will be deleted?
-. tie point limit of 16,000 will become equal to 12,000 tie point limits (25% reduction) or 8,000 (50% reduction) or 4,000 (75% reduction)
-. tie point limit from any original setting process (16,000 or 12,000 or 10,000 etc) will strictly reduce to a default tie point setting value (remain only equal to the result of the 4,000 setting as the default)
-. percentage reduction from the result of the original tie point (from 36,000,000 tie points become 18,000,000 tie points (-50%) or maybe 9,000,000 (-75%)

Maybe, through understanding the nature of this filtering option will shed light on reducing the problem and dealing with our machine's performance.

Many thanks.

2

General / Re: Append intended for part of a chunk

« on: June 28, 2023, 03:31:38 PM »

Hi Alexey,

thank you for the explanation.

Regards.

3

General / Re: Cleaning Tiled Model (Remove Floating blobs)

« on: June 24, 2023, 01:29:11 PM »

Hi Mitchgalea,

have you finally been able to clean your tiled model? Which approach do you use?
I experienced a similar problem as yours, and I use depth maps based on tiled model generation having the same parameter setting as well.

Thank you.

4

General / Re: Append intended for part of a chunk

« on: June 24, 2023, 11:53:42 AM »

Hi Alexey,

for the same question, is it possible with the so-called "import" process?

Thank you.

5

General / Re: Append intended for part of a chunk

« on: June 23, 2023, 02:39:28 PM »

Hello Alexey,

thank you for the assistance. Okay, I will try to do that.

6

General / Tiled model surface number – “dense cloud” and “depth maps” based options

« on: June 21, 2023, 12:19:31 PM »

Has anybody an idea whether a “tiled model” based on dense cloud and depth maps has a similar default number? For example, if the high default in the dense-cloud-based option shows 1 million (number in a bracket) and it applies also to the depth maps based.

Because in the computer I use, the High – Medium – Low default parameter using the dense cloud as the base indicates the number as well (in a bracket), but there is no number in the depth maps option.

If it is different, the high number of faces in the depth maps option will produce a different number of faces for the dense-cloud-based option. It could be a higher number or less number. If it is so, which one will produce more faces and therefore will have more possibility to have TDR problem?

The information and answer provided here may be useful to anticipate that problem.

Thank you for the enlightenment and for reading this question.

7

General / Relation between depth maps and dense cloud to dense cloud cleaning process

« on: June 21, 2023, 12:15:25 PM »

Hi everybody, I have tried to search for the information in this forum but I have not been lucky enough to find the answer.

When we proceed “build dense cloud” job, I think it automatically produces depth maps as well, doesn’t it? The question is: after cleaning or improving our dense cloud, either with gradual selection or filter by confidence options, it will clean or improve our depth maps also, or not?

If the improvement process of dense cloud would also take effect on the depth maps, it means when we process the tiled model based on depth maps, it uses the improved depth maps as the reference or source. That is a benefit, I guess. Or, the build-tiled model process will produce other (new) depth maps by disregarding the existing depth maps?

In this case, we put the same parameter of depth maps (on the tiled model source) and enabled the keep depth maps option (on the advance preference).

Thank you for the enlightenment and for reading this question.

8

General / Append intended for part of a chunk

« on: June 21, 2023, 12:09:23 PM »

Is anybody know the possibility to append only part of the chunks? It means we can append “tiled model” or “dense cloud” or other items only, instead of appending the entire chunks, in order to complete our backup file.

I have 2 files that are located in two different drives. One is in the computer for the working process, and another is in an external drive for backup. Every time the process is finished it will be copied (append) to the external drive as backup. For example, if the “build dense” cloud process has been finished, the entire chunk will be saved or copied in the external. The process continues with “build mesh” and “tiled model” by using the file in the computer. When finished, the entire chunks (photo alignment, dense clouds, mesh, and tiled model) the whole chunk will be saved again in the external drive as a new folder. This step requires a longer time and occupies more space in the external.

When I will execute append, I saw an option to choose to append “tiled model” and the other items as well.

Can we utilize this option for completing our backup in external drive? Or is this option intended for another purpose? If it is not possible, is there an alternative way or step to do it? So, we should not copy or append the entire chunks?

Thank you for reading this query and for the enlightenment given.

9

General / Re: TDR problem – Nvidia? How to handle it?

« on: May 31, 2023, 01:03:40 PM »

Hello Alexey, thank you for the prompt response.

My project is not an aerial survey but a kind of close-range scanning. The total picture I have is around 86,000. I split them into 7 smaller chunks and a final chunk, with enough overlapping each of them, having respectively 23,571 pictures, 24,683 pictures, 19,565 pictures, 14,595 pictures, 13,316 pictures, 15,651 pictures, 49,756 pictures, and the final chunk contains all of them together. It is quite consuming time and patience. Since this is a massive heritage building built in 700 CE, therefore involving many reliefs and details, making it more difficult to split them into more chunks than the above stated.

The recently running process is the one with 15,615 pictures with a data size exceeding 450 GB.

The computer we use has an Intel Xeon processor, and Quadro graphic card (P-2000), and is supported with 256 GB RAM. Agisoft Metashape Pro 1.8.3 is installed. We have not yet moved to 2.0 because of certain considerations and the excellence of this version in assisting us becomes one of the utmost reasons.

In my case, I need to analyze the antique measurement and proportion produced by the reconstruction of the target model. In other words, I need to have a solid and detailed 3D model, not the points, so I can research it and print it later.

Regarding your suggestions, what will happen if I reduced the face count set to 100 to 200 million? Would it reduce the quality drastically? By default, the parameter setting of the High count states 2,483,000,000 faces, the Medium has 828,000,000 faces, and the Low one has 268,000,000 faces. I kept the high faces as default and just started to create another one by trying the Medium faces setting (828,000,000), but still applied dense cloud as the source.

And what will happen if I use Depth Maps as the source instead of Dense Cloud?

I believe the other chunks above containing more pictures will have more faces.

What do you mean by generating a tiled model instead of the build-mesh? Do you mean directly going to, or generating the tiled model, and skipping or without processing the build-mesh? Is that okay? Because I thought the tiled model is the following or respective step after the build mesh process.
If that is possible, it will be an excellent way out for reducing the exhausting process.

And please Alexey, would you please provide me the guidance for the detailed steps, including the parameter settings?

Many thanks and regards,
Adiwan.

10

General / TDR problem – Nvidia? How to handle it?

« on: May 30, 2023, 03:59:56 PM »

Hello Everybody, I need inputs regarding TDR.

When opening the result of a Build Mesh, I got a TDR error. What is this and how to handle it?

Build Mesh parameters applied are Dense Cloud as the source, Arbitrary (3D), surface count High (2,483,000,000), interpolated, and calculate vertex colors.
The other parameters such as Dense Cloud quality, Depth Filter, Strict Volumetrict, and reuse Depth Maps options were greyed out.
It is the Build Mesh command from workflow, not through the Batch process. This procedure created a 3D Model Mesh with approximately 1,973,000,000 faces but experienced a TDR error when trying to open it.

The Dense cloud itself was produced with Ultrahigh quality with mild filtering and then filtered by confidence, which was reduced by around 50% points to become 9,546,404,254 points (points after filtering).

If this problem is due to the machine's performance, is changing Build Mesh parameters become the way out? For example, just use the Low surface count option (or lowered it down by using custom surface count).
Or it is better to choose Depth Maps as the source and Low surface count instead of Dense Cloud and High surface count options.

Many Thanks and regards.

11

General / Re: Pause and the Consequences

« on: May 16, 2023, 03:14:45 PM »

Thank you Olihar for the information. It means, after executing "pause" command, i can open directly and start another project, or should I wait until the project is truly paused or stopped? Because the other users may not be able to wait for hours until it really stops or paused. They usually want to use it directly, after a certain number of minutes.

Regarding the version, we may be considering the newer version later, because more users here still utilizing it as well.

12

General / Pause and the Consequences

« on: May 12, 2023, 06:47:53 PM »

Hi Admins and Metashapers,

we have a problem with the number of computer users, but only have a computer dedicated to Agisoft Metashape tasks. Therefore, would you please give us some information regarding the “pause” command? The questions are:

1.   What is going to happen if we pause a process if another person has an urgent need to use Agisoft as well, for a certain of hours? Is the in-hold process will be in a secure and good condition to be continued later? Or it will be useless because the “pause” damages the entire process?

The recently running process that would be held is a Build Dense Cloud of a single-quite big chunk having 35,000 + photos. The data size exceeds 350 GB.
What I know is the pause command takes certain hours to completely respond a big data. I have experienced a response that took 2 days+ and failed to be resumed. Also for Build Dense Cloud. It was a chunk with 50,000+ cameras for approximately 651 GB.
I am also not sure whether Agisoft Metashape Pro could accept other jobs, while the currently running job is in-hold position, as mentioned in https://www.agisoft.com/forum/index.php?topic=8965.msg42040#msg42040.

2.   If it is truly safe to pause a task, in which percentages or maybe steps are suggested to execute it? Because Build Dense Cloud has several steps like generating depth maps, and so on; photo alignment also has matching photos, camera locations, etc., etc.

Almost forget, the computer we use has Xeon processor, and Quadro graphic card, and is supported with 256 GB RAM. Agisoft Metashape Pro 1.8.3 is installed.

I hope this forum would give us enlightenment about it because I tried to search for further information but have not been lucky enough to find it.

Many thanks,

13

General / Re: Merging chunks (can't merge the dense point clouds)

« on: November 18, 2022, 08:57:21 PM »

Hello warrenburton,

would you please share me the steps in making chunks alignment? I have tried it but it seems to be unsuccessful.

Many thanks and regards,
AS.

Hello

Please help.

2 questions that I had asked a while ago but no answers:

1. I created 4 chunks with images, tie points, depth maps and dense clouds. I have aligned the chunks and all chunks have R next to them. When I merge the chunks and select dense cloud in the merge option, the resultant merged chunk only has the cameras and tie points and no depth maps or dense clouds (even though I had selected 'dense clouds' in the merge chunks options. How do I ensure that all 4 dense clouds are merged into one? I would like to use the new merged chunk to be able to build a DEM and orthomosaic.

2. To try something different I exported dense clouds for each chunk as DCP files and imported only the DCP files into a chunk. I then deselected the 'set as default' and selected all 4 dense clouds and merged them. The result only gets to 95% before it finishes yet there is no merged dense cloud to be see at all. Where could it be? With this merged dense cloud I would like to create a DEM and orthomosaic.

Thank you.

14

General / Re: Why is it "Selecting pairs..." when I only want to merge chunks?

« on: November 18, 2022, 08:43:18 PM »

Just want to add another information and queries:

1. For the recently chunks alignment process that has not finished yet, I do not checked "keep key point" in the preference tab, will it be the problem?
2. The remaining disk-space is 279 GB, I hope it will sufficient because I read in an article in this forum that unsufficient space may create problem as well.
3. Instead of aligning chunks with point or camera based option, can I employ append function as mentioned in the following article, but both chunks are still in tie points https://agisoft.freshdesk.com/support/solutions/articles/31000157701-manual-chunk-alignment-standard-edition-

Many thanks and regards,
AS

15

General / Re: Why is it "Selecting pairs..." when I only want to merge chunks?

« on: November 18, 2022, 07:11:05 PM »

Hello Alexey, jenkinsm, and other members....

I have experienced similar problem. I have executed chunks alignment process, for 2 chunks, with approximately 35% overlaps one to another, since 3 days ago but the process has not finished yet until today.
The first chunk contains 10,847 photos, 1 component, with 3,476,786 points (120,000/12,000) and the second one has 15,651 photos with 27,254,766 points (160,000/16,000). Both chunks showed excellent results, but still in tie points, because I want to merge both chunks with point based alignment.
Except KPL and TPL abovementioned, the other parameters are the same, which are: Highest accuracy, Generic preselection, Source, Adaptive camera, No mask, and No stationary

Batch process was employed to handle the chunks alignment, and the parameter was: Point based, Highest accuracy, KPL 160,000, No mask, and Generic preselection.
The first step, which was saving project run about 1 hour, the following steps or Detecting points took 1 days 16 hours, and the third or Selecting pairs as the next process has took 1 days 14 hours until now. So in total it has 3 days.

When I was watching the end phase of the Selecting pair process, this morning, actually, it has reached 100% within 1 days and 04 hours 17 minutes.

Meaning that the job has around 7 hours without changing to another processing steps. Because as an alignment chunks process, in total, when selecting pairs process finished, the "green-bar" indicator reached only around 50%.

The questions are:
1. After selecting pair process, how many steps are remained for this alignment chunks process?
2. Which steps require more or longer period of time?
3. Considering the case I have, do you have any suggestions? Should I abbort the operation or wait?

I have 2 options in my mind when abborting or cancelling this process:
1. Try another procedure of chunk alignment with camera based, or
2. Add more photos to the chunks having more photos, and run photo alignment with Keeping Key Point on the preference + without reset current alignment.
Since I aligned the photos of those both chunks without Keeping Key Points option checked (off) 2 weeks ago, it is okay if I turned on the preference now.... or should I realigned again both chunks from scratch before aligning them. It took abot 40 days for both chunks.

I hope someone who success or expert in doing chunks alignment and merging proceess would kindly share the tips and tricks. Your enlightenment would be highly appreciated.

Many thanks and regards,
AS