Can provide data for a test case？

[Mr-ZhuJun]

Hello, your work is very new, and I would like to ask if you can provide the data of a test case, so that you can see the effect more intuitively.

[JiawangBian]

Thanks for your question. We have already provided one test case here https://github.com/ActiveVisionLab/porf/blob/df5ef77f34c7314ac1adc14b836dd4a8931ea746/scripts/train_sift_dtu.sh#L7

You can download the dataset and run this script.

[Mr-ZhuJun]

hellow，I generated a model based on the data, and the result of the model is completely wrong. Can you help me take a look?

python train.py --mode validate_mesh --conf confs/dtu_sift_porf.conf --case scan37

[Mr-ZhuJun]

[JiawangBian]

You need to load the model weights by adding "--is_continue".

[Mr-ZhuJun]

thanks for your answer！ link: https://github.com/ActiveVisionLab/porf/blob/07b26c5962f6942b43b7d08a75873071cba10380/train.py#L390 I didn't find this option"--is_continue". Could you please give me more details?

[JiawangBian]

Sorry. We find that the current version doesn't include it to load the pre-trained model. This is because the current version is only designed for pose optimisation. Here, the mesh visualization is used for debugging. The mesh should be automatically saved after training (https://github.com/ActiveVisionLab/porf/blob/07b26c5962f6942b43b7d08a75873071cba10380/train.py#L215)

If you need high-quality mesh results as shown in the paper. You should save the optimised pose, and use it to train a Voxurf model.