Max Zhaoshuo Li 李赵硕

Hi @VitorGuizilini-TRI The training and evaluation scripts are provided in the scripts folder.

Hi @yanziwei I only had 1 gpu so the code was not written for DPP by default. You need to wrap the model into DPP training pipeline following this tutorial:...

Is the Table 3 in the paper reporting 3px error?

Hi @Torment123 Thanks for reaching out. 1: Yes, occlusion information is used to train STTR 2-3: I have not trained on dataset without occlusion information. The scene flow result is...

Hi @cp319 Thank you for your interest in the project. I am not able to reproduce your error. Maybe you have something off during your evaluation? Please see the result...

@cp319 Yes. I am using the same file. Here is the set of arguments I used for your reference. ``` --epochs 1 --batch_size 1 --dataset kitti2015 --dataset_directory /data/Data/KITTI/2015 --checkpoint inference...

I tried the following: ``` pip install git+https://github.com/agoose77/K3D-jupyter#chore-use-hatch --force-reinstall pip install ipywidgets==7.7.1 jupyterlab-widgets==1.1 --force-reinstall ``` But still seeing this error ``` [Open Browser Console for more detailed log - Double...

Hi @xuduo18311199384 You can see the functions [here](https://github.com/NVlabs/neuralangelo/blob/main/projects/neuralangelo/scripts/convert_data_to_json.py#L190) of how to obtain the above values!

Hi @xuduo18311199384 Great results! The noise of reconstruction generally comes from the homogenously color-ed background (eg., pure white) or unobserved regions. If you consider these undesirable, you can filter out...

Hi @pierreparfait01 One option is to set the checkpoint frequency high enough. But this will require many disk storage.