How to run this code on our own dataset.?

[DRAhmadFaraz]

@samarth-robo
Respected Sir,
As this code run on ur dataset of "7-Scenes" now if we want to run this code on our custom images then How would this code will able to evaluate and predict the results.?

as in Dataset of "7-Scenes" It needs two things as an input.
1 is "RGB image"
2 is "frame-000001.pose.txt" files having 4x4 pose matrix.

so in our own custom images, Can you please guide me how to get this 4x4 pose matrix "frame-000001.pose.txt" from every RGB images.?

I will be thankful to you and waiting for your kind response.

Regards

[samarth-robo]

Hi @DRAhmadFaraz, sorry for the late reply. Somehow I missed the notification for your first comment.

There are a couple of ways to get the camera pose:

1. You can record a video (preferably with an RGB-D camera like a Kinect), and then use some SLAM algorithm get the camera trajectory + map of the scene. This is exactly how the authors of 7 Scenes created their dataset, they used a SLAM algorithm called KinectFusion. You can read their papers to get more details.
2. Kinect does not work outdoors. So if you want an outdoors dataset, you might have to use stereo cameras or a Lidar sensor. They have their own specialized SLAM algorithms. I don't know the state-of-the-art, but I think LSD-SLAM or Stereo DSO should work well.
3. If you have an unordered set of images from the scene, instead of video streams, you can use Structure from Motion (e.g. the implementation from mapillary) to estimate the camera poses for every image.

[DRAhmadFaraz]

@samarth-robo Thanx a lot for helping me, I have a set of RGB images so I think point # 3 would be helpful for me.

can you please guide me which tool is best for Structure from Motion (SFM) to extract the information same as in a format as shown in your dataset text files corresponding to every image "frame-000001.pose.txt" exactly needed for your algorithm as an input. as shown below.

9.3075085e-001 1.6080184e-002 -3.6457023e-001 -1.4914335e-001
7.6722316e-002 9.6786028e-001 2.3863889e-001 -1.3233277e-001
3.5676569e-001 -2.5012836e-001 8.9978862e-001 1.8095869e-001
0.0000000e+000 0.0000000e+000 0.0000000e+000 1.0000000e+000

So that I might able to train your algorithm on my own data-set captured from simple RGB camera..

I will be thankful to you.

[samarth-robo]

I mentioned an SfM tool in my previous comment. Sorry but I'm not very familiar with the area, and don't know other resources.

On Sun, Apr 28, 2019 at 17:43 Ahmad Faraz Khan ***@***.***> wrote: @samarth-robo <https://github.com/samarth-robo> Thanx a lot for helping me, I have a set of RGB images so I think point # 3 would be helpful for me. can you please guide me which tool is best for Structure from Motion (SFM) to extract the information same as in a format as shown in your dataset text files corresponding to every image "frame-000001.pose.txt" exactly needed for your algorithm as an input. as shown below. 9.3075085e-001 1.6080184e-002 -3.6457023e-001 -1.4914335e-001 7.6722316e-002 9.6786028e-001 2.3863889e-001 -1.3233277e-001 3.5676569e-001 -2.5012836e-001 8.9978862e-001 1.8095869e-001 0.0000000e+000 0.0000000e+000 0.0000000e+000 1.0000000e+000 So that I might able to train your algorithm on my own data-set captured from simple RGB camera.. I will be thankful to you. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#16 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAVXKRQO7RYMIEA67I54NL3PSWIFTANCNFSM4HHONW5A> .

-- Samarth Brahmbhatt https://samarth-robo.github.io/

[DRAhmadFaraz]

Ok sir, I will try this method,

Thanx a lot..