CloudCompare

**Andrea Alessi** · 2016-05-04, 04:33 PM

Hi Mateusz,

You should be able to see your mesh the same way you do when enabling/disabling point clouds. Since you are not seeing anything, it could be anything from an old driver to something more serious, so i need more information.

Could you please share with me at least the log files ( https://www.3dflow.net/technology/documents/3df-zephyr-tutorials/how-to-locate-3df-zephyr-log-files/ ) or even better, the whole dataset and zep file ?

Thank you,

Andrea

**MateuszB** · 2016-05-04, 05:23 PM

Hi Andrea,
Thanks for reply, I`ve sended log file and bigger description of the project.

Mateusz

**Andrea Alessi** · 2016-05-04, 07:30 PM

Thank you for your followup.

I have sent you via email some more specific information regarding the dataset, however it looks like this is a bug that appears for some unfortunate reason in that specific dataset.

We're currently working on it! Thank you again for helping us making Zephyr a better product! I'll keep you updated both via email and forum!

**Andrea Alessi** · 2016-05-05, 01:58 PM

Hi gagamel,

thank you for reporting this problem. You're actually right, it looks like there are some rare cases in which for some reason (currently under investigation) Zephyr fails to merge cameras clusters. This doesn't happen in most cases/datasets but as you demonstrated it can happen.

You did nothing wrong on your part

For the time being, if you see that the orientation phase fails merging clusters, i'm afraid the only solution at the moment is to re-run the first phase (to know how many camera clusters have been detected, look for

[18:02:49] Total Clusters found: 2

in the logfile)

I apologize for the incovenience and thank you for your bug report!

Andrea

**Andrea Alessi** · 2016-05-11, 08:03 PM

just a small update: we have fixed this bug in our internal development version!

We expect to release the new version hopefully at the end of next week (around 20th may) or the end of the month (around 27th may) alongside many other improvements!

Thank you!

**gagamel1** · 2016-05-12, 03:28 PM

Hi Andrea

that are good News. Thank you very much.

**techworkstudio** · 2016-06-02, 08:55 PM

Hi everyone, I had a few questions and was hoping that the community here would help resolve my problems.

I have a small team and we're in the very early stages of creating a video game for the Vive, and we wanted to be able to use this software to help with the modeling and such for the game, but i'm just trying to learn the software right now, and I've been encountering a couple of issues that I was hoping you could assist me with.

1. On my first trial I used a tissue box, the first time I was using a cannon digital camera, and what I think happened was that the resolution of the images was not good enough, so even on exhaustive I was only getting maybe 75-80% of the cameras, and the dense point cloud was spotty in some areas and perfect in others. I then proceeded to start using my Nikon D60 camera, which made all the difference in the world.

2. On the second test I used a bucket, and this time I made smaller movements around the object and took pictures as close to the same area as possible when taking at different heights. This time the software accepted all 48/48 cameras which was exciting, the dense point cloud looked good about 80% around but on one side of the bucket there were almost no points, I wasn't sure what that was going to look like, but after generating the mesh I found that about 60% of the bucket was modeled perfectly, but the other side looked like it was melted with acid.

3. After this I tried to model an Xbox 360, and this is where I started encountering some new problems. It seems that the software was having trouble recognizing the flat sides of the Xbox where it's all 1 color and there is very little if no texture at all. I assume the software looks for changes in texture and color, so I can only guess that modeling objects with flat single colored sides to it wasn't going to work.

4. My next trial was a tissue box, this time I had a really interesting experience. After taking all my pictures and setting everything up into the software, I was happy to find that all 46/46 pictures were accepted, so I continued on generating the dense point cloud. Again the dense point cloud looked great on 3/4 sides and the 4th side had maybe 4 or 5 total points that was it.... also I was having an issue where the software was showing as if one angle of the box had no cameras at all.... My guess is that since the box is the same on all 4 sides both in design, texture, and color, that the software clumped pictures from 2 sides together thinking that they were the same side, and when I generated the mesh and texture I found that three sides looked PERFECT, the tissue box was an exact replica, but the one side of the box just wasn't there.

5. After all of that I decided it was time to try modeling a person, I knew that it was going to be more difficult to do this, but I wanted to jump right in.... my first attempt was a complete failure... my cameras were not overlapping enough I didn't give myself enough space around the person to walk around and keep the same distances or similar distances all around... so I went at it a second time, but this time I was extra extra careful and detail oriented with each and every step... I set up my camera on a tripod, I went outside to a large square area in my driveway and position by position walked around a center marked spot marking the position of the legs on the tripod to get as close to 100% overlap as I possibly could. My first height was about waist high, next was about chest high, and the third height was maybe 2 or 3 inches over the persons head.

I was very excited to see how this was going to generate in the software so I went inside and started working. I set the software to human body, and put the settings on deep just to be safe. It took roughly 35 minutes, but afterwards I was happy to see that I had literally gotten almost 100% overlap on all three heights, my angles were perfect and the distance between each picture looked amazing. Next I generated the dense point cloud and the results were mixed... the software didn't generate almost anything from the waist down, but EVERYTHING else looked amazing. Things got weird from here... after generating the mesh I was already disturbed, the software had almost duplicated her face and mushed it together into some weird cartoon like almost scary blob, and her legs were completely missing all the way down to her feet which were modeled. After texturing the mesh things looked BETTER, like her upper body excluding her head was modeled absolutely flawlessly, but her head was still so screwed up, her lower body was missing, and it was generating too much of the ground around her.

So now to my questions.

1. For small to medium sized objects, what is an ideal amount of pictures to take per height and how many different height levels should I take pictures at?

2. How can I get the software to stop modeling the environment around the object as well as the table or surface the object is sitting on?

3. If I have an object that has 2 or more similar sides, how can I get the software to recognize that there are four full complete sides and prevent it from cutting one off all together?

4. When taking pictures can I spin the object around instead of moving the camera around the object... like If I had small flat surface on a track that could spin that I could put my object on top of?

5. When trying to generate a model of a person what is the best way to do it? How many different heights should I take pictures at? Should I take a set of pictures at ground level? How many pictures should I take at each height?

I will post pictures of the trials and errors I have gone through soon, I took quite a few pictures of the mishaps and such.

I apologize for the plethora of questions, but this is going to be a HUGE part of our project and I want to get it perfect.

Thanks in advance for any help, it is all appreciated!

**techworkstudio** · 2016-06-02, 09:05 PM

Here are all of the pictures I took of my progress periodically throughout the trials. I wish I had more, but I hope this is enough.

http://imageshack.com/a/img921/6907/bHfLIX.png - Another view of the most recent person model attempt
http://imageshack.com/a/img924/7741/G52ddV.png - Overhead Dense point cloud of the most recent person model attempt
http://imageshack.com/a/img924/8859/6H11pP.png -Melted half of the bucket trial
http://imageshack.com/a/img922/3486/TbyT3p.png - zoomed out view of most recent person model with camera angles
http://imageshack.com/a/img923/8391/QT3fd0.png - Results of the most recent person model
http://imageshack.com/a/img924/3880/WFFGSX.png - Tissue box front... can't seem to download the back, but the other side you can't see is just missing
http://imageshack.com/a/img922/3742/Hmx6RR.png - Bucket Dense Cloud

**Andrea Alessi** · 2016-06-20, 02:31 PM

Quick followup: this bug has been fixed in the just released version 2.600!

CloudCompare

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment

Comment