mhamilton723
evaluation code missing
Hi, thanks for your work!
Evaluation code seems missing. Any reason for that?
To reproduce the main table of the paper I should take the checkpoints, train the probes, and code an evaluation loop, right?
What would you do if you had to reproduce table 1?
Hi, I am also interested in the evaluation code to reproduce Table 1 results!
Hi all, you guys can look into my implementation here: https://github.com/andrehuang/loftup/blob/main/eval_seg.py for the reproduction of semantic segmentation. For depth estimation, I recommend using the depth and normal experiments from Probe3D which is in my opinion a better practice for evaluation.
