Fu-En Wang

Yes, it will be the same. Because eventually, I will still normalize the sphere to a unit sphere. So you can actually ignore this parameters. The reason why I add...

@topcoder613 I have merge the equirectangular to cubemap and cuebmap to equirectangular utility. Please see this. https://github.com/fuenwang/PanoramaUtility

You mean coordinate on equirectangular? https://github.com/fuenwang/Equirec2Perspec/blob/master/Equirec2Perspec.py#L71 The lon and lat is the coordinate.

The lon and lat is the coordinate of h x w which is consistent with the perspective images. So you can just use lon[0, 0] and lat[0, 0] ...... to...

@LeeBC2298 I think it's indeed cosface. The output of the classifier (https://github.com/MuggleWang/CosFace_pytorch/blob/master/layer.py#L43) is the cosine similarity with margin. And directly optimizing the cosine similarity with cross entropy loss is equal...

I just saw it now haha. What do you mean? The orientation our model predict can be directly converted to rotation_y. So we will use it to solve T.

@jinfagang https://github.com/fuenwang/3D-BoundingBox/blob/master/Eval.py#L70 This line is the rotation_y the model predict.

 The theta in that link is the theta(red) in this figure, so it's already the global rotation of object, not local angle and it is the angle you need...

1. Yes, camera sees local angle. 2. Actually I'm not familiar with alpha either haha, but as far as I know, we can also infer local angle by it. The...

But his code got much more stars than mine haha. My code can converge, you can check this closed issue https://github.com/fuenwang/3D-BoundingBox/issues/4 My final angle and dimension error is also in...