IOError: [Errno 2] No such file or directory: 'data/IR2/reconstruction.json'

[kkkarthik6]

Hi, I am trying to reconstruct a scene and it has good feature matches and all but is giving me following error

2018-02-22 11:24:35,152 DEBUG: Robust matching time : 0.000420093536377s 2018-02-22 11:24:35,152 DEBUG: Full matching 125 / 146, time: 0.129931926727s 2018-02-22 11:24:35,221 DEBUG: 04.jpg - 05.jpg has 262 candidate matches 2018-02-22 11:24:35,222 DEBUG: Robust matching time : 0.000317096710205s 2018-02-22 11:24:35,222 DEBUG: Full matching 237 / 262, time: 0.123347043991s 2018-02-22 11:24:35,695 INFO: reading features 2018-02-22 11:24:35,742 DEBUG: Merging features onto tracks 2018-02-22 11:24:35,780 DEBUG: Good tracks: 1068 2018-02-22 11:24:36,181 INFO: Starting incremental reconstruction 2018-02-22 11:24:37,473 INFO: 0 partial reconstructions in total. Traceback (most recent call last): File "/home/karthik/spring/OpenSFM/OpenSfM/bin/opensfm", line 34, in command.run(args) File "/home/karthik/spring/OpenSFM/OpenSfM/opensfm/commands/mesh.py", line 22, in run reconstructions = data.load_reconstruction() File "/home/karthik/spring/OpenSFM/OpenSfM/opensfm/dataset.py", line 371, in load_reconstruction with open(self.__reconstruction_file(filename)) as fin: IOError: [Errno 2] No such file or directory: 'data/IR2/reconstruction.json'

Kindly let me know what issue can this be.

[paulinus]

@kkkarthik6 from the logs it seems that it has indeed some matches but still did not manage to reconstruct it. This happens mostly when there is not enough parallax on the images. That is when the camera is not displaced between shots. Can you share the images you are using?

[kkkarthik6]

Hello Pau,

Thank you for your comment. Here are the images. Kindly let me know if there is any work around to reconstruct these images. I also have GCP for these and I wanted to reconstruct the path based on path and the features obtained.

Best regards Vamsi Karthik Vadlamani, Research Assistant, University of New Mexico. +1 571-385-8979

On Thu, Mar 1, 2018 at 9:06 AM, Pau Gargallo notifications@github.com wrote:

@kkkarthik6 https://github.com/kkkarthik6 from the logs it seems that it has indeed some matches but still did not manage to reconstruct it. This happens mostly when there is not enough parallax on the images. That is when the camera is not displaced between shots. Can you share the images you are using?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mapillary/OpenSfM/issues/293#issuecomment-369640476, or mute the thread https://github.com/notifications/unsubscribe-auth/ACTimRCb44FDD_Y18GqdPyItYqy-tL4Gks5taBxygaJpZM4SPxc6 .

[paulinus]

@kkkarthik6, i can't find the images on your message. If you attached them to an email, then probably github has removed them. Can you paste them to the issue on github?

[kkkarthik6]

Sure

On Thu, Mar 1, 2018 at 9:32 AM Pau Gargallo notifications@github.com wrote:

@kkkarthik6 https://github.com/kkkarthik6, i can't find the images on your message. If you attached them to an email, then probably github has removed them. Can you paste them to the issue on github?

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mapillary/OpenSfM/issues/293#issuecomment-369649669, or mute the thread https://github.com/notifications/unsubscribe-auth/ACTimadVIGys-6fveMVeVZZt_XDHrTotks5taCKzgaJpZM4SPxc6 .

-- Best regards Vamsi Karthik Vadlamani, Research Assistant, University of New Mexico. +1 571-385-8979

[kkkarthik6]

Here are the images.

[paulinus]

@kkkarthik6, I'm afraid these images are not suitable for SfM. There is very little texture that can be used to infer correspondences. The matches you are getting are most likely all in the overlayed numbers. (those have no parallax and thus you get no reconstruction)

[paulinus]

What are those btw? they look interesting

[kkkarthik6]

They are extracted from thermal imaging camera, these are Infrared Images. Actually I have extracted few features from these on my other experiment to prepare a compass using these images. Any suggestion to improve these images?

[paulinus]

The most important would be to mask out the numbers. You can do that by cropping the images or by adding binary images to masks folder. The mask for image 01.jpg should be in the file masks/01.jpg.png as in the Berlin example.

Even after masking the overlay, there is really little texture for finding point correspondences. So the reconstruction will not work. You would need to have less blurry images to get more correspondences.

[kkkarthik6]

Sure , I will let you know if there is any impovment in getting this done.

On Thu, Mar 1, 2018 at 10:26 AM Pau Gargallo notifications@github.com wrote:

The most important would be to mask out the numbers. You can do that by cropping the images or by adding binary images to masks folder. The mask for image 01.jpg should be in the file masks/01.jpg.png as in the Berlin example.

Even after masking the overlay, there is really little texture for finding the point correspondences. So the reconstruction will not work. You would need to have less blurry images to get more correspondences.

— You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub https://github.com/mapillary/OpenSfM/issues/293#issuecomment-369665593, or mute the thread https://github.com/notifications/unsubscribe-auth/ACTimZJzALNSnvLgiC8YlWgMKlhc-qAtks5taC8BgaJpZM4SPxc6 .

-- Best regards Vamsi Karthik Vadlamani, Research Assistant, University of New Mexico. +1 571-385-8979

[YanNoun]

No answer for a while, closing.