Forum

Show Posts

This section allows you to view all posts made by this member. Note that you can only see posts made in areas you currently have access to.


Messages - Steve003

Pages: 1 ... 6 7 [8] 9 10 ... 12
106
Hi,
VERY IMPORTANT QUESTION !
I read that Metashape works with the exif data in order to perform its magic.
If I am hitting a stills capture button in a video playback or editing prog, capturing stills from the video, will those have that exif data the same as a stills camera would generate ?
would such images captured that way work in Agisoft ?

I need to capture stills at moments in the video I know I need them, two stills might be 3 seconds apart, using Agisoft  inbuilt image capture every 10th frame will see far too many stills compared to what is required for 60% overlap. If I end up with 10x as many images as required it will make the prog struggle if not fail ? Some of those are bound to be between keyframes, keyframes give the best image usually, if it captured every lets say 5th keyframe that might be better.

agreed a slow shutter speed means a slower pan, but a dull and poorly lit museum needs a slow shutter speed if I am to have good depth of field.

Also the Sony AX100E I have seems to have an inbuilt inability to pan without causing jerkiness, as I have tried it in sunshine on everything from 1/50 to 1/600.

My subject matter pan starts off showing a load of exhibits in the background then becomes in seconds just my subject then the ceiling creeps in then the floor, all that extra background stuff seems to me a distraction, likely to cause failure. I am familiar with mask from model, but if the background isn't enough in the sequence to create a decent mesh, editing it out can see one faced with a poorly defined mush to start with.

I have had failures where I am convinced pre editing in photoshop would give a better result.

Steve003

107
General / Re: Turning object around- chunks
« on: August 10, 2019, 03:08:22 AM »
apart from a circular additive paintbrush, a straight line polygon tool as in photoshop would make masking in Agisoft far easier, as such as I need to use that tool or the ability to use quickmask, black as foreground colour (becomes pink) hold shift down and click with paintbrush to get straight line sections that way, I have to use photoshop for masking.

Steve

108
please again, anyone ????

109
Hi,
as I was panning slowly as my 4K camcorder seems to give jerky video on normal panning despite trying the fastest card and best settings, and its £1400 and full manual control, every 10th frame would give far too many images.
I am interested to know, if a subject is 2 meter long, and I am standing 2 metres away, how many seconds panning from start to finish might one need to do to get a suitable number of frames if every 10th frame is used. I was taking maybe 16 seconds to do so. I later did a faster scan maybe 8 secs as another test.

I would have to on this occasion manually edit out the backgrounds on some of the frames in photoshop, as it goes from a shot of the subject and a lot of background in the museum when starting the scan to just the subject.

Steve

110
Hi Kiesel, I can see mention of import video in the pdf manual, it says:-
Import Video...imports video in a form of frame sequence and saves the extracted frames as images.
I panned really slowly to avoid jerky stills, but I wouldnt want e.g. 1 still per second as I would have thousands of stills. I wonder what extraction options there are. I also will need to mask the background out as its in a museum and if it works on what it sees it will emit smoke signals !
page 17 of the metashape pdf manual doesnt have anything on video and use of, (p17 is titled reconstruction parameters) I have searched on the term video and found nothing, ditto 4k.
I envisaged another video editing prog to get exactly the stills I want from it.

Steve

111
Hi,
a camera takes a still image perfect if shutter greater than focal length.
A camcorder when panning  is a different thing, must the stills be as perfect for photoscan or Metashape to work.
Would a 4K camcorder save taking loads of photos, would we have to extract the stills or can Agisoft work with video ?

detail of what I tried...Using an AX100E 4K camcorder by Sony, panning can look a bit jittery, even very slow panning, using 1/50 second 0db f6.8 0density filter, lighting dictated this if using 0Db gain, would a faster a shutter speed help ? but then the depth of field would suffer, or would it, are f numbers a different thing on camcorders ? It was held on wide angle throughout the sequences.

Steve

112
anyone please ?

113
General / Re: Aligning chunks
« on: July 30, 2019, 02:56:07 PM »
anyone ?

114
General / Re: Aligning chunks
« on: July 22, 2019, 09:42:39 PM »
I also would love to know more about Align chunks and see no one has explained such in your question.
Merge chunks simply brings the photo sets together. If one chunk has three faces of a building , face 1 2 and 3 and another chunk faces 3 and 4 is one face in common enough ?
I often have to photo an object then turn it upside down to do the underside, and such sees one side re-photographed, so agisoft needs to align that side with the same side in chunk 1. My METHOD D drawn up with Alexie's help doesnt mention using align chunks at all, the merged chunks all have masks from model so I guess agisoft is not aligning chunks but merely seeing a masked photo with faces 3 and 4, and another with faces 1 2 and 3 and bringing them together when making the dense point cloud., however my projects such as this all fail.

They work if the object isn't moved. so far anything I have done fails.

would align chunks stop that failure I wonder.

Steve003

115
Hi,
Quote
you can only reuse depth maps if the ones you already have were generated at the same 'quality' setting that you are trying to reuse them at.
but I had gone workflow>build dense cloud and used high, then saw that I couldnt select reuse depth map when selecting high in build mesh. Depth maps get made in the 'Build dense cloud' process I am told, so if I wish to use qlty high for mesh I must use qlty high for dense cloud stage, I am told.

Quote
anyway if you are generating mesh from depth maps then there's no need to build dense cloud.
I dont see an option to build depth map.  In workflow after I have merged chunks, then selected all thumbs and reset camera alignment, the next step is Build Dense Cloud. I use qlty 'high' and depth filtering 'mild' as per Alexei advice.

is my workflow list missing showing me a stage or option ?

Steve003

116
Hi,
when photographing an object e.g. a long thin box each side 30mm x 350mm placed on the ground and walking around it, first sub sequence top and both sides and ends, then turn it upside down to do the underside, can that sub sequence be the underside and one side and both ends , the ends being so narrow they couldnt assist the process that well in alignment ?
Will Photoscan or Metashape recognise the one side and marry it to the first chunks side, or does it need both sides and underside photographing in that second sub sequence ?

I ask as I had a 3 side sub-sequence (top and both sides and both ends) then an underside and one side and both ends sub-sequence and merge chunks then dense point cloud and mesh build failed to create all four sides with major omissions. (Alexei has project as Leaf spring looking at, though I simplify the description to 'box' just to establish the principle)

Also Merge chunks and Align chunks, what is difference as my method D sheet doesnt feature Align chunks at all, and Merge chunks puts all images together into one chunk it says.

Thanks

Steve

117
General / Re: Agisoft filling in bolt and rivet holes I need to see.
« on: July 20, 2019, 12:16:40 PM »
Hi JMR, thank you,
in order to check my new positioning of this fix hole step, as you say 'NO MORE BUILD DENSE CLOUD', here is METHOD D V5m (drawn up with Alexei assistance) before (step 2.4.1) and after altering its location to a new step location 3.1.5 after the last 'build dense cloud' of the method D steps.

METHOD D v5m (a work in progress method sheet, NOT DECLARED WORKABLE AS YET but if anyone uses it and it works please let me know)
(for when objects are moved e.g. to photo the underside. We call the photos for each object/background combination a sub-sequence) so 1 chunk per sub-sequence.
Up to here we have aligned images in each chunk.
2.2.3) Workflow>Build dense cloud
2.2.6) when done go Workflow>Build mesh and source Depth maps and in advanced select reuse depth.
DON’T FORGET Repeat dense point cloud and mesh creation for each chunk.
2.3) cut the unwanted geometry mesh that is not related to the object of interest and generate masks from the model in each chunk.
2.3.5) go File>Import>masks from Models, replacement, all cameras, repeat for each chunk.
2.4) Workflow>Merge Chunks
2.4.1) To fix holes filling in on brake shoe model:-
- remove (reset) all the masks from the chunk,
- mask the six holes in the objects on the very first image from the set (that is looking from the top on the object),
- run Build Mesh (in Metashape 1.5, of course) using Depth Maps as a source, same quality as the existing depth maps in the project (enabling Reuse Depth option) and using Strict Volumetric masks.

2.4.3) double click Merged chunks in list at left , select all thumbnails, right click apply ‘reset camera alignment ’.
2.4.5) Workflow>Align Images  use Apply Masks to Key Points option. ( chunks are merged only to have all the images with the respective masks in the single chunk. Also select ’reset current alignment’ to sort out any problems, Also though not vital untick ‘Adaptive Camera Model Fitting.’
3.1) Workflow>Build Dense Cloud,  for quality choose high or ultra high, depth filtering Mild, tick Calculate Point Colours. Alexei used LOW QUALITY in this step  to fix the failed brake shoe. This was for speed not quality.
3.2) Workflow>Build Mesh, source Depth maps, open advanced options and tick ‘reuse depth’, Quality choose the same as was chosen in Step 3.1   and using Strict Volumetric masks.

METHOD D v6m provisional NOT DECLARED WORKABLE AS YET but if anyone uses it and it works please let me know)....await JMR input.
2.2.3) Workflow>Build dense cloud
2.2.6) when done go Workflow>Build mesh and source Depth maps and in advanced select reuse depth.
DON’T FORGET Repeat dense point cloud and mesh creation for each chunk.
2.3) cut the unwanted geometry mesh that is not related to the object of interest and generate masks from the model in each chunk.
2.3.5) go File>Import>masks from Models, replacement, all cameras, repeat for each chunk.
2.4) Workflow>Merge Chunks
2.4.3) double click Merged chunks in list at left , select all thumbnails, right click apply ‘reset camera alignment ’.
2.4.5) Workflow>Align Images  use Apply Masks to Key Points option. ( chunks are merged only to have all the images with the respective masks in the single chunk. Also select ’reset current alignment’ to sort out any problems, Also though not vital untick ‘Adaptive Camera Model Fitting.’
3.1) Workflow>Build Dense Cloud,  for quality choose high or ultra high, depth filtering Mild, tick Calculate Point Colours. Alexei used LOW QUALITY in this step  to fix the failed brake shoe. This was for speed not quality.
3.1.5) To fix holes filling in on brake shoe model:-
- remove (reset) all the masks from the chunk,
- mask the six holes in the objects on the very first image from the set (that is looking from the top on the object),
- run Build Mesh (in Metashape 1.5, of course) using Depth Maps as a source, same quality as the existing depth maps in the project (enabling Reuse Depth option) and using Strict Volumetric masks.

3.2) Workflow>Build Mesh, source Depth maps, open advanced options and tick ‘reuse depth’, Quality choose the same as was chosen in Step 3.1   and using Strict Volumetric masks.

JMR, is this now where it should be if I understand you correctly.

many thanks

Steve

118
General / Re: metashape wont show jpg in add photo
« on: July 12, 2019, 08:47:18 PM »
Hi, found the culprit. me :-[

I had navigated to a folder of almost same name, had raw file canon cr2 in it !

however add photos now ok,

HOWEVER...metashape fails to align , 4 failed, yet photoscan did all of them  :o

see email 11 july  15:46 where pscan 1.2.6 failed 12 and metashape 20 same settings. good qlty pics as well
Nothing to do with blame on pics as far as I can see. if it was then pscan would fail same or even more.

These this time being long and thin had d.o.f fall off, unavoidable yet pscan no probs.

Regards

Steve

119
General / metashape wont show jpg in add photo
« on: July 12, 2019, 07:29:28 PM »
Hi,
having firstly tried photoscan 1.2.6 and browsed to the jpg folder, added the photos there and aligned them (all ok)  I now try metashape to see if I get any failures, but it shows no jpg in the folder yet they are still there. relaunch it and try again, even choose jpg from file type list lower right but its as if the folder is empty.

unable to add photos to project.

aim it at another folder, also shows it with no files to pick from, its gone blind as it were !

Steve

120
Hi Alexei,
Quote
If each chunk represent the fixed object-to-background relative orientation, then the images of the same sub-set should properly align, providing that there's sufficient overlap between the images.
quite right and I have sent you the project file and images, as you will see they overlap well. 1 chunk per object fixed relative to background.

Item sitting on the gridded board on concrete driveway in a yard, I moved around it. one of many projects and photo sessions like that. No chance of a background image or two or 3 representing entire background. the object predominates the photo so of the area of it photographed it should make a model, but if they fail to align I cant even go ahead and make dense point cloud/depth map and then make the model from which to edit the background from, then import mask from.

pics that dont align at the outset, 24 out of 54 failing,  thats 'game over'.

Quote
Masking from model, if applied to Entire Workspace, would generate masks for the images from the certain chunk basing on the model from the same chunk.
Are you saying I should choose 'Entire Workspace'  as opposed to all cameras when importing masks from model for chunk 1 ?

I havent got as far as that yet because I havent even made dense point cloud or mask,  I dont have decent models to remove background from yet. I have only aligned and with so many unaligned I cannot progress.

Quote
It is not possible to generate the masks for the images from the "Chunk A" basing on the model from the "Chunk B".
Agree wouldnt have crossed my mind.

first step is to align pics having added photos to project and they wont align, I can accept 1 or 2 failures but not 24 out of 54.

I then try photoscan 1.2.6 and get almost half the failures. I have sent an email with differences.

Should not metashape better photoscan ? The settings clearly make a difference.

I used:-
Accuracy: HIGH
Pair preselection: DISABLED
Key point limit: 40,000
Tie Point Limit: 10,000
Constrain features by mask: greyed out unselectable
Adaptive Camera Model Fitting :  ON


try my images in photoscan 1.2.6 to see what I mean.

Regards

Steve

Pages: 1 ... 6 7 [8] 9 10 ... 12