gap with results in papers

[BenjaminGit001]

Hi, Thanks for your impressive paper and code. I tried this repo to reproduce this performance, I followed all instruction and trained on300w-lp use train.py without change any parameters, then evaluate on AFLW2000 using test.py and results as below:

me: Yaw: 3.9897, Pitch: 5.0923, Roll: 3.6405, MAE: 4.2408 yours: Yaw: 3.63 , Pitch: 4.91 Roll: 3.37 , MAE: 3.97

Is there any other tricks or changes should be apply for reproduce your results?

[thohemp]

Hi,

the reported results are generated with the same code and setup from this repo. You can try adjusting the learning rate or scheduler (e.g. decreasing the decay from .5 to .1) to check if this improves the results.

There are multiple factors like different hardware, model initialization and random sample loading that can lead to slightly different results. Especially the last two points will induce randomness to each training.

[thohemp]

@BenjaminGit001 I updated the default training settings. This should help you to achieve the reported results more easily (or in my trainings even surpass them). Let me know if you still struggle after using them.

[zxcvbnm12316]

Hi, Thanks for your impressive paper and code. I tried this repo to reproduce this performance, I followed all instruction and trained on300w-lp use train.py without change any parameters, then evaluate on AFLW2000 using test.py and results as below:

me: Yaw: 3.9897, Pitch: 5.0923, Roll: 3.6405, MAE: 4.2408 yours: Yaw: 3.63 , Pitch: 4.91 Roll: 3.37 , MAE: 3.97

Is there any other tricks or changes should be apply for reproduce your results?

Hello, I retrained the model and got the ‘’epoch.tar‘’ file. The pressurized file contains the ‘’data.pkl‘’ file, and the instance file shows the need for.‘’pth‘’ file. How do I get the.pth_ file for snapshots,thank you。

Required in the document： python test.py --batch_size 64 \ --dataset AFLW2000 \ --data_dir datasets/AFLW2000 \ --filename_list datasets/AFLW2000/files.txt \ --snapshot output/snapshots/1.pth \ --show_viz False

I trained to get the files _epoch_30 .tar Unzip the file, which contains the following three files data data.pkl vwesion