Topics

1

General / Can someone upload test photos of persons full body

« on: April 27, 2015, 02:59:50 PM »

Just wondering if someone could upload a image set of a full body scan. Im wanting to test out a few things but I dont have any good images to work off of. Not looking for anything in particular, just good photos from head to toe so I can run it through Agisoft.

Thanks in advance

2

General / NVIDIA Tesla Benchmarks

« on: April 18, 2015, 04:21:56 AM »

So I got tired of ray tracing taking forever and picked up a pair of Tesla cards. Figured I would give them a run at Agisoft and share some informaiton on them:

Nvidia Tesla M2090
6GB DDR5
225W TDP
Average power usage with Agisoft 150-180w
Idle power usage 20-30w
Idle temps 30-47C
Average temps under load 70-90C (passive cooling right now)
ECC Memory State turned ON

I am getting a lot of throttling use to temps right now, the case was not designed for passive cooling co-processors.  I have a few ideas I will be testing out this weekend to try adn get the load temps to under 70C all the time.  Ocne I do I will run the tests again. After multiples tests with the Agisoft benchmark data I got the following average samples/sec

Medium Reconstruction 500-600 million samples/sec

I think I might be able to squeeze a little more out of the cards by turning off ECC memory state, which will raise the memory bandwith. I will test this out tomorrow after I work on getting the cards cooler.

All in all I am very happy with these cards and their performance. They are under permorning the $300 GTX 780 cards I have, but at the price poitn I got them for, $75 each, they are simply amazing.

Just wanted to share since there has been a few other interested in Tesla/Phi chard performance. Check back later this weekend for more results

3

General / 1080p and 4K Video --> Point Cloud --> Mesh

« on: March 05, 2015, 06:13:56 PM »

Ive been playing around with a bunch of non traditional ways of creating meshes and point clouds in Agisoft and wanted to share some recent results. Photogrammetry is all just pictures to points to create point clouds, but what is video and can we create pointclouds using individual frames from video?

What I have found is that it is possible and can be faster to capture and record a object with video to process later than taking photos. The nice thing about video is that depending on the camera you can capturing upwards of 60+ frames per second and can really use anywhere from 50-90% of the frames. I say 50-90% because depending on lighting and your shutter speed you might get some blur ever few frames, but either way you will still generally have 500+ useable frames per minute of video recorded. This is nice when time is limited and you need to quickly get in and out of an area.

I have two examples below that I created using two completely different methods of video to mesh design. This first example is from a cemetery. I took a close up, 3-5ft away, 1080p video using a Sony camera. Due to the resolution you will need to keep the camera pretty close in order to get decent detail. The entire video was 30 seconds long to create this model and ended up being 250 pictures once I pulled frames.

https://sketchfab.com/models/337263c53bdc4f908f64c7ffa7b74b61

The second example is a test that I did using Google Earth and Camtasia to screen capture what I was looking at on my monitor. Due to the fact that I have 4K monitors I was able to record my screen at 4K and then process the video the same way I did the model above. The total video was 60 seconds long and I pulled 625 frames to process the entire project.

https://sketchfab.com/models/c29a84f14a3f45cdaf1166cd15756a6e

These scans are obviously not the most accurate ones out there but if time is limited and you have access to decent video cameras this process might work for what you need. Ill be testing a few other projects over the next few weeks and post up more results as I get them.

Hope you all enjoy

4

General / 1,000+ Photo Project Workflow

« on: February 27, 2015, 06:45:30 PM »

I have seen a few people on here talk about extremely large dataset they have worked with on projects, upwards of 1,000+ photos, and was wondering what the normal workflow was for a project this big? I would guess something like this would be processes in different chunks with multiple overlapping pictures in each chunk. Im just a little lost when it comes to stitching everything together. I know with scanners like the Faro x130 you can use Scene to stitch everything together, but with the scans from Agisoft whats the best way to merge everything back into one large file?

5

General / Low light capture recommendations

« on: February 24, 2015, 06:56:18 PM »

I have a project that I am wanting to tackle where there will be no light, it’s a .25 mile tunnel with graffiti. I am wondering what’s the best way to tackle this sort of project. Should I look into multiple flood lights and try to illuminate once section at a time? Just wondering if anyone else has done similar work and if so how sucessful they were with their process.

6

General / Best GPU for the money

« on: February 04, 2015, 10:12:59 PM »

So I have just built a new setup for my home to replace my old 1366 workstation. This will be used for 3D printing, 3D mesh design, photogrammetry, CAD (2D/3D), photoshop and a ton of other things. I have a Quadro K5000 that will be used for most of my work, but I am wanting to fit the system with 3 other GPUs in order to play around with larger Agisoft creations. Which cards would you recommend as the best bang for the buck? From looking around the forums it looks like the 780/780ti would be the best, but just want to make sure before I pull the trigger on 3 more cards.

Thanks

PS - here is the new system spec.

Asrock X99 WS Motherboard
Xeon E5-2643 6-Core CPU
128GB (8x16GB) DDR4 2133
1500W PSU
Quadro K5000 GPU
Corsair 540 Air Case
256 Samsung 850 Pro SSD (Boot drive)
2TB Corsair MX100 SSD Storage (4x512GB drives)