I got a problem at using camera matrices in your dataset.
I rendered multi-view images and depth maps by using the metadata the author of 'R2N2' have provided. In addition, I checked out that the camera matrices which are generated from the metadata, matched those in your dataset. After then, I tried point unprojection by using the multi-view depth maps with the corresponding camera matrices. However, unrpojected points at each view show a partial shape of an object but those at multi-views does not represent a 3D object shape at all. I think that something's wrong with the camera matrices.
Sorry to trouble you, but could you mind checking this out and telling me what the problem it is?
Hello, there.
I got a problem at using camera matrices in your dataset.
I rendered multi-view images and depth maps by using the metadata the author of 'R2N2' have provided. In addition, I checked out that the camera matrices which are generated from the metadata, matched those in your dataset. After then, I tried point unprojection by using the multi-view depth maps with the corresponding camera matrices. However, unrpojected points at each view show a partial shape of an object but those at multi-views does not represent a 3D object shape at all. I think that something's wrong with the camera matrices. Sorry to trouble you, but could you mind checking this out and telling me what the problem it is?