The test results seems to be inconsistent with the paper

[runningpp]

I tried your code in Ubuntu20.04 cuda 11.8 pytorch2.0.1
I have got the results:

GTA.mp4

but the corresponding results in your paper is much better obviously, as shown in the following：

I want to know if my result is reasonable。

[River-Zhang]

Thanks for your feedback! That's weird. Here are our results showing in Meshlab:

Could you kindly display the command you executed along with the output produced by the code? Additionally, I would appreciate it if you could execute the testing script. This will allow me to ascertain whether the test results on CAPE are within expectation.

The only explanation I can think of currently is that there are discrepancies between our environments. We didn't adopt PyTorch versions beyond 2.0 due to compatibility issues with certain features.

[runningpp]

Thanks for your feedback! That's weird. Here are our results showing in Meshlab:

Could you kindly display the command you executed along with the output produced by the code? Additionally, I would appreciate it if you could execute the testing script. This will allow me to ascertain whether the test results on CAPE are within expectation.

The only explanation I can think of currently is that there are discrepancies between our environments. We didn't adopt PyTorch versions beyond 2.0 due to compatibility issues with certain features.

The command I used:
''
python -m apps.infer -cfg ./configs/GTA.yaml -gpu 0 -in_dir ./examples -out_dir ./results -loop_smpl 100 -loop_cloth 200 -hps_type pixie
"
Note that: I change the GTA.yaml: normal_path: "./data/ckpt/normal.ckpt" to "./data/ckpt/Normal-epoch=19-v5.ckpt", because the normal mode you provided in CKPT is Normal-epoch=19-v5.ckpt.

The test results in CAPE are:

{'cape-easy-NC': 0.03472820296883583,
'cape-easy-chamfer': 0.6505321264266968,
'cape-easy-p2s': 0.6074833273887634,
'cape-hard-NC': 0.04288707673549652,
'cape-hard-chamfer': 0.9164005517959595,
'cape-hard-p2s': 0.8493930101394653}

The I change the the environments to " cuda11.6 pytorch1.13.0", I found the results between the two environments are almost the same. The following is the results on pyotorch 1.13.0:
{'cape-easy-NC': 0.03472774103283882,
'cape-easy-chamfer': 0.6541368961334229,
'cape-easy-p2s': 0.6101600527763367,
'cape-hard-NC': 0.04288462921977043,
'cape-hard-chamfer': 0.9143836498260498,
'cape-hard-p2s': 0.8439348340034485}

Maybe it is not caused by the environments. The results you showed have much better detailed surface, can you tell me how to get your shown results?