[Question]: Evaluation

[Vaiduryasses]

Your question

Hi @chenzhaiyu ，
when i use ckpt_best.pth you submitted, it turns out that :
`2025-04-26 11:00:38,229 - PaCo - INFO - Starting reconstruction...

Loading weights from ./ckpt/ckpt-best.pth...

Failed to reconstruct 00852735

Failed to reconstruct 00618377

Failed to reconstruct 00152622

Failed to reconstruct 00500275

[Timeout] Task 2327 (model 00268571) timed out.

2025-04-26 11:10:53,006 - PaCo - INFO - Generating evaluation stats...

Running loop for <function _hausdorff_distance_single_file at 0x7fbe4b67f250> with 2366 calls on 4 subprocess workers

Running loop for <function _chamfer_distance_single_file at 0x7fbe4b67f1c0> with 2366 calls on 4 subprocess workers

Running loop for <function _normal_consistency at 0x7fbe4b67f2e0> with 2366 calls on 4 subprocess workers

2025-04-26 11:15:55,566 - PaCo - INFO - Test pipeline completed.`

then i check the evaluation.csv , it's all like this :
`in mesh,ref mesh,Hausdorff dist new-ref,Hausdorff dist ref-new,Hausdorff dist,Chamfer dist(1: no input),Normal consistency(0: no input)

./output/default/reconstructed/00000952.obj,data/abc/eval/ref_obj/00000952.obj,1.0,1.0,1.0,1.0,0.0

./output/default/reconstructed/00003544.obj,data/abc/eval/ref_obj/00003544.obj,1.0,1.0,1.0,1.0,0.0

./output/default/reconstructed/00003963.obj,data/abc/eval/ref_obj/00003963.obj,1.0,1.0,1.0,1.0,0.0

./output/default/reconstructed/00004055.obj,data/abc/eval/ref_obj/00004055.obj,1.0,1.0,1.0,1.0,0.0
`
How can i deal with this?

[chenzhaiyu]

Hi @Vaiduryasses, a few of the 2k+ samples would fail, hence PolyFit’s FR > 0. Please see Supplementary Section C.2 for details on handling failures. Note that the provided checkpoint actually slightly outperforms the results reported in the paper.