The experiment about Fig. 8 in the paper

[XinhuaZhang]

Hi Chen-Hsuan,

I trained a barf on the fern scene. I am not sure if the results below agree with SfM. In figure 8 I don't see the distance between the estimated poses and the ground truth poses. Is this normal?

[chenhsuanlin]

Hi @XinhuaZhang, the poses here are the raw optimized poses and are variable up to a 3D similarity transformation. The poses in Fig 8 are Procrustes aligned. Please see the Evaluation criteria section in the paper for details.

[XinhuaZhang]

Thanks