Forum

Author Topic: Generating Dense Points: Longer it runs the more time is left  (Read 3038 times)

B_Free42

  • Newbie
  • *
  • Posts: 10
    • View Profile
Generating Dense Points: Longer it runs the more time is left
« on: October 26, 2016, 04:46:30 AM »
Photoscan has been generating the dense point cloud for about 18 hours now. It is currently 2016-10-25 20:19:00 local time and the last entry in the Photoscan log is from this morning at 3:08. It says "filtering depth maps..."

Earlier this afternoon, the top progress bar indicated there was about 24 hours left.  Now, six hours later or so, it's telling me there is over 31 hours left. The longer I let Photoscan run the more time it says it has remaining. See attached screenshot from just now.

Is the program stuck in some kind of loop? Will it ever finish executing?  At this point should I cancel the dense point cloud generation? It's run for over 30 hours already, I'd hate to give up on it now. What's going on? Am I doing something wrong?

The input is 142 tif photos - tif photos that were converted from RAW files. Each one is 6000 x 4000 pixels and about 70 Mb in size. This is the second time I've tried generating a model with tif files - the first time I got memory error after some 30 hours.  Is there something I do to change this?

Here some excerpts from the log. Starting from when I first loaded the photos.

Quote
2016-10-24 09:10:25 Agisoft PhotoScan Version: 1.2.6 build 2834 (64 bit)
2016-10-24 09:10:25 OpenGL Vendor: NVIDIA Corporation
2016-10-24 09:10:25 OpenGL Renderer: GeForce GTX 560 Ti/PCIe/SSE2
2016-10-24 09:10:25 OpenGL Version: 4.5.0 NVIDIA 372.90
2016-10-24 09:10:25 Maximum Texture Size: 16384
2016-10-24 09:10:25 Quad Buffered Stereo: not enabled
2016-10-24 09:10:25 ARB_vertex_buffer_object: supported
2016-10-24 09:11:28 Loading photos...
2016-10-24 09:11:37 Finished processing in 8.942 sec (exit code 1)
2016-10-24 09:12:05 Detecting points...
2016-10-24 09:12:10 photo 1: 39936 points
2016-10-24 09:12:15 photo 2: 39936 points
...
2016-10-24 09:24:44 photo 142: 39935 points
2016-10-24 09:24:44 points detected in 758.89 sec
2016-10-24 09:24:44 Matching points...
2016-10-24 10:21:42 3218536 matches found in 3417.64 sec
2016-10-24 10:21:42 matches combined in 0.522 sec
2016-10-24 10:23:49 matches filtered in 126.273 sec
2016-10-24 10:23:49 setting point indices... 157705 done in 0.091 sec
2016-10-24 10:23:49 generated 157705 tie points, 4.19292 average projections
2016-10-24 10:23:49 removed 139456 multiple indices
2016-10-24 10:23:49 removed 152 tracks
2016-10-24 10:23:49 finished matching in 4303.51 sec
2016-10-24 10:23:49 selected 147632 tracks out of 157553 in 0.027 sec
2016-10-24 10:23:49 Estimating camera locations...
2016-10-24 10:23:49 processing block: 46 photos
2016-10-24 10:23:49 pair 66 and 67: 1 robust from 3073
2016-10-24 10:23:49 pair 9 and 115: 0 robust from 1968
2016-10-24 10:23:49 pair 5 and 6: 764 robust from 1920
...
2016-10-24 10:25:14 adding 6185 points, 4416 far (0.883977 threshold), 3998 inaccurate, 123 invisible, 132 weak
2016-10-24 10:25:14 optimized in 20.838 seconds
2016-10-24 10:25:15 coordinates applied in 0 sec
2016-10-24 10:25:15 Finished processing in 4389.16 sec (exit code 1)
...
2016-10-24 10:39:40 loaded project in 0.047 sec
2016-10-24 10:39:40 Finished processing in 0.047 sec (exit code 1)
2016-10-24 10:41:29 Using device: GeForce GTX 560 Ti, 8 compute units, 2048 MB global memory
2016-10-24 10:41:29   max work group size 1024
2016-10-24 10:41:29   max work item sizes [1024, 1024, 64]
2016-10-24 10:41:29   max mem alloc size 512 MB
2016-10-24 10:41:29 Initializing...
2016-10-24 10:41:29 initializing...
2016-10-24 10:41:29 selected 142 cameras from 142 in 0.37 sec
2016-10-24 10:41:29 Loading photos...
2016-10-24 10:43:36 loaded photos in 126.312 seconds
2016-10-24 10:43:36 Reconstructing depth...
2016-10-24 10:43:37 [GPU] estimating 4804x2588x1088 disparity using 801x863x8u tiles, offset -72
2016-10-24 10:43:38 [CPU] estimating 5360x3750x672 disparity using 1072x938x8u tiles, offset -426
2016-10-24 10:44:24 timings: rectify: 1.453 disparity: 45.745 borders: 0.264 filter: 1.188 fill: 0
...
2016-10-25 00:37:54 [GPU] estimating 4150x7351x768 disparity using 830x919x8u tiles, offset -240
2016-10-25 00:39:07 timings: rectify: 1.978 disparity: 71.401 borders: 0.317 filter: 1.586 fill: 0
2016-10-25 00:39:10 [GPU] estimating 5802x3733x1184 disparity using 829x934x8u tiles, offset -728
2016-10-25 00:41:08 timings: rectify: 1.16 disparity: 117.034 borders: 0.148 filter: 1.124 fill: 0
2016-10-25 00:41:10 [GPU] estimating 5705x3862x992 disparity using 951x773x8u tiles, offset -511
2016-10-25 00:42:17 timings: rectify: 1.184 disparity: 66.066 borders: 0.179 filter: 1.145 fill: 0
2016-10-25 00:42:20 [GPU] estimating 5574x3982x608 disparity using 929x797x8u tiles, offset -147
2016-10-25 00:43:05 timings: rectify: 1.175 disparity: 44.117 borders: 0.393 filter: 1.193 fill: 0
2016-10-25 00:43:08 [GPU] estimating 5490x3967x352 disparity using 915x794x8u tiles, offset 0
2016-10-25 00:43:39 timings: rectify: 1.164 disparity: 29.591 borders: 0.739 filter: 1.275 fill: 0
2016-10-25 00:43:50 finished depth reconstruction in 50414.8 seconds
2016-10-25 00:43:50 Device 1 performance: 100.008 million samples/sec (CPU)
2016-10-25 00:43:50 Device 2 performance: 264.061 million samples/sec (GeForce GTX 560 Ti)
2016-10-25 00:43:50 Total performance: 364.069 million samples/sec
2016-10-25 00:43:51 Generating dense point cloud...
2016-10-25 00:44:04 selected 142 cameras in 13.507 sec
2016-10-25 00:44:04 working volume: 11898x5618x6405
2016-10-25 00:44:04 tiles: 3x1x2
2016-10-25 00:44:04 selected 97 cameras
2016-10-25 00:44:04 preloading data... done in 45.04 sec
2016-10-25 00:44:49 filtering depth maps... done in 7978.17 sec
2016-10-25 02:57:53 preloading data... done in 415.672 sec
2016-10-25 03:04:48 accumulating data... done in 103.957 sec
2016-10-25 03:06:33 building point cloud... done in 0.516 sec
2016-10-25 03:06:36 selected 141 cameras
2016-10-25 03:06:36 preloading data... done in 85.767 sec
2016-10-25 03:08:01 filtering depth maps...

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: Generating Dense Points: Longer it runs the more time is left
« Reply #1 on: October 26, 2016, 11:39:37 AM »
Hello B_Free42,

What is the quality setting that you've used for the dense cloud generation and what are the computer specifications? Also have you checked the memory consumption and CPU activity in Task Manager?
Best regards,
Alexey Pasumansky,
Agisoft LLC

B_Free42

  • Newbie
  • *
  • Posts: 10
    • View Profile
Re: Generating Dense Points: Longer it runs the more time is left
« Reply #2 on: October 26, 2016, 03:41:30 PM »
Hi Alexy,
I used the highest settings for Dense Point Cloud Generation. My computer specifications are:

Intel Core i7-2600K @3.40 GHz, 3801 Mhz, 4 Cores, 8 Logical Processors
RAM: 24GB
64 bit, Windows 10
One GPU: NVidia GeForce GTX 560 Ti

As you can see from the Task Manager screen shots the CPU useage is in the 3-5% range while the memory useage has been consistently up in the 97% range.

Also, the program is still running this morning which is good, although the time left has increased to 49 hours.

Thank you very much for your help, Alexy!

Alexey Pasumansky

  • Agisoft Technical Support
  • Hero Member
  • *****
  • Posts: 14813
    • View Profile
Re: Generating Dense Points: Longer it runs the more time is left
« Reply #3 on: October 26, 2016, 03:54:48 PM »
Hello B_Free42,

Looks like there's insufficient RAM, so PhotoScan is using swap and is transferring the data between the swap and RAM.
Actually, not sure if it would be possible to finish the process in a reasonable time, as the progress is not near to the end.
Best regards,
Alexey Pasumansky,
Agisoft LLC