The panoramic view is not stiched well (first and last image are looking up or down with non-zero elevation) in each scan

[lukewenMX]

The generated panoramic view from the skybox images which are downloaded by configuring the filename = matterport_skybox_images when runing ./download_mp.py program, is not stiched well by runing the downsize_skybox.py.(matterport3D simulator: https://github.com/peteanderson80/Matterport3DSimulator#simulator-api). That is, the <pano.id>_skybox_small.jpg is not an accurate panoramic view especially for the first and last images (one is looking up and one is looking down with an non-zero elevation) . An instance has been shown in below, please give some comments and suggestions, thanks.

[lukewenMX]

I am wondering whether it is the dataset problem? since the original skybox images are not always with zero elevation.

[Yoark]

@lukewenMX Hi, are you able to obtain panoramic image now? I am trying to generate them as well

[lukewenMX]

@Yoark Yeah. Actually, you can use the sandbox images to generate them by only stiching the images with index from 1-4 (modify the downsize_skybox.py in the Matterport3DSimulator/scripts folder). or the simplest but raw way is screenshot from the generated sandbox_small images.

[Yoark]

@lukewenMX Hi, Luke, I also found this https://gist.github.com/ronghanghu/d250f3a997135c667b114674fc12edae and https://github.com/yindaz/PanoBasic for visualization of panorama. the panobasic one provides a real panorama rather than just stitching images.

[nivesh48]

Hi, @Yoark and @lukewenMX. Even I'm trying to figure out how to stitch perfect panoramic images from the available skybox images. Any leads on this ?

[Yoark]

Hi, @Yoark and @lukewenMX. Even I'm trying to figure out how to stitch perfect panoramic images from the available skybox images. Any leads on this ?

you can check the panobasic link for panoramic from skybox images.

[nivesh48]

Thanks, @Yoark. I'm currently looking at it and for my needs, I need to stitch the panorama from undistorted images. As suggested in the panobasic repo, we need to calculate view direction vectors from extrinsic so. Even if you are into this keep me posted if you find anything helpful.