[OpenDroneMap-users] Tips and tricks for model re-runs

[Alex Mandel]

It's not the data set. Most of us have been hitting this bug with the
case of JPG not working and needing them to be jpg.

https://github.com/OpenDroneMap/OpenDroneMap/issues/33

I'll check that data set and update the names if they are still in caps
in git.

Thanks,
Alex

On 03/11/2015 12:11 AM, Frederic Muller wrote:
> Hi!
> 
> Just installed ODM and trying to have it run first. I'm using the images
> GitHub of the project and specifically ran on COPR images.
> 
> Unfortunately it has failed twice with the same error message which is
> "Error in OdmTexturing:
> Error in OdmTexturing:
> Failed to read image:
> '/home/fred/Documents/odm_data/copr//IMG_0031.jpg'
> 
> I looked at the image and it opens fine, but maybe the problem comes
> because of the double /?
> 
> Or is copr data set not ready?
> 
> Thank you.
> 
> Fred
> 
> 
> On 03/10/2015 11:25 PM, Alex Mandel wrote:
>>
>> I added a full gcp list to the copr data set. It runs but I have yet to
>> verify how well it did in real world terms.
>>
>> Feel free to test with that.
>>
>> Thanks,
>> Alex
> 
> On 03/05/2015 07:11 PM, Stephen Mather wrote:
>> Hi Ivan,
>>
>> Boruszyn now has GCPs associated with it:
>>
>> https://github.com/merkato/odm_boruszyn_kap
>>
>> PUWG1992
>> 340577 547925 70 691 3498 img_4851.jpg
>> 340610 548131 71 1248 2921 img_4854.jpg
>> 340649 548176 70 1078 2312 img_4912.jpg
>>
>> And all these GCPs are valid when processed. Boruszyn does not make ideal
>> orthos, due to the geometry difficulties of a KAP flight, but it's such an
>> amazing dataset.
>>
>> Somewhere I have a couple more GCPs for Langley, but I can't find
> them... .
>> I know Aaron Racicot, the originator of that dataset, is working both on
>> GCPs and EXIF camera positions for that dataset. He's nearly done with
>> that, so stay tuned.
>>
>> Also, there's a new graceful-exit branch (thanks to Spotscale) to test:
>>
>> https://github.com/OpenDroneMap/OpenDroneMap/tree/graceful-exit-gcp
>>
>> Please give this a test on datasets that have no EXIF and no GCPs.
>>
>> Thanks,
>> Best,
>> Steve
>>
>>
>>
>>
>> On Wed, Mar 4, 2015 at 8:03 AM, Ivan Lizarazo <ivan.lizarazo at gmail.com>
>> wrote:
>>
>>> Nice work, Steve. It's really helpful.
>>>
>>> I ran ODM with the Langley dataset using this gcp_list.txt (taken
> from the
>>> wiki):
>>>
>>> *WGS84 UTM 10N*
>>> *544256.7 5320919.9 5 3044 2622 img_0525.jpg*
>>> *544157.7 5320899.2 5 4193 1552 img_0585.jpg*
>>> *544033.4 5320876.0 5 1606 2763 img_0690.jpg*
>>>
>>> After a while, ODM quit the georeferencing job.  This is the last part of
>>> the odm_georeferencing_log.txt:
>>>
>>> *Reading following GCPs from file:*
>>> *x_: 544257 y_: 5.32092e+06 z_: 5 pixelX_: 2622 pixelY_: 3044 image:
>>> img_0525.jpg*
>>> *x_: 544158 y_: 5.3209e+06 z_: 5 pixelX_: 1552 pixelY_: 4193 image:
>>> img_0585.jpg*
>>> *x_: 544033 y_: 5.32088e+06 z_: 5 pixelX_: 2763 pixelY_: 1606 image:
>>> img_0690.jpg*
>>>
>>> *The calculated easting offset for the georeferenced system: 544149*
>>> *The calculated northing offset for the georeferenced system: 5.3209e+06*
>>>
>>> *Recalculated GCPs with offset:*
>>> *x_: 107.7 y_: 21.9 z_: 5*
>>> *x_: 8.7 y_: 1.2 z_: 5*
>>> *x_: -115.6 y_: -22 z_: 5*
>>>
>>> *Reading mesh file
>>>
> /home/ivan/odm_data/langley/reconstruction-with-image-size-2400-results/odm_texturing/odm_textured_model.obj*
>>>
>>> *Successfully loaded
>>>
> /home/ivan/odm_data/langley/reconstruction-with-image-size-2400-results/odm_texturing/odm_textured_model.obj.*
>>> *Position in model for gcp 3: x=13.1883 y=-2.29253 z=-15.2447*
>>> *Error in Georef:*
>>> *Less than 3 GCPs have correspondences in the generated model.*
>>>
>>> I understand that more than 3 GCP are needed to succeed.  Do you  have
>>> more GCP available for the  Landley dataset?
>>>
>>> Could anybody provide a GCP list for any of the ODM sample datasets?
>>>
>>> Best regards,
>>>
>>> Ivan
>>>
>>> On Sun, Mar 1, 2015 at 8:32 PM, Stephen Mather <stephen at
> smathermather.com>
>>> wrote:
>>>
>>>> Hi All,
>>>>
>>>> I was playing today with the Langley dataset, working on the
>>>> georeferencing / orthophoto tests, and hadn't loaded enough (useable)
>>>> points in my gcp file for the georeferencing / orthophoto generation.
>>>>
>>>> So, I added some more ground control points to
> odm_georeferencing_log.txt
>>>> and wanted to rerun, but not everything. Since the latest step we
> can start
>>>> from (at this time) is pmvs, here was the command I ran so that I
> wouldn't
>>>> have to rerun all the matching and bundler:
>>>>
>>>> run.pl --resize-to 3000 --start-with pmvs
>>>>
>>>> It completed in 30 minutes instead of 8 hours, which was refreshingly
>>>> short, and reran PMVS (point cloud model creation), meshing, mesh
>>>> texturing, and orthophoto generation.
>>>>
>>>> Hope this helps,
>>>> Best,
>>>> Steve