Hello,
First, congrats on your amazing work and thanks for releasing the code. I was wondering how to reconstruct unknown shapes at test time. In particular, Table 1. Would it be possible to share the code for the same? If not, can you please briefly mention whether you optimise the joint loss. Do you minimise the weighted sum of individual loss as follows?
$$\$ \mathcal{L}_{total} = w_{fine} \mathcal{L}_{fine} + w_{coarse} \mathcal{L}_{coarse} + w_{kld} \mathcal{L}_{kld} \$$$
Finally, would you be willing to accept a Pull Request so that in case someone else would like to use these results in the future?
Thank you for your time
Hello,
First, congrats on your amazing work and thanks for releasing the code. I was wondering how to reconstruct unknown shapes at test time. In particular, Table 1. Would it be possible to share the code for the same? If not, can you please briefly mention whether you optimise the joint loss. Do you minimise the weighted sum of individual loss as follows?
Finally, would you be willing to accept a Pull Request so that in case someone else would like to use these results in the future?
Thank you for your time