Request for Demo and Qualitative Results of Your Implementation

[mansooreh1]

Hello.
I would greatly appreciate it if you could share a demo of your implementation with me. Additionally, if possible, I would like to request qualitative results implementations showcasing the effectiveness of your method.

Your insights and any additional materials you can provide would be extremely valuable for my understanding and future research directions. Thank you for your time and consideration.

[Sirui-Xu]

Hi, would you mind clarifying what you mean by "qualitative results implementations"? Are you referring to the visualizations? If so, we do provide the code here: https://github.com/Sirui-Xu/InterDiff/tree/main/interdiff/render.

[mansooreh1]

"Thank you for your answer, yes exactly, I mean visualization. Thanks for clarifying how I should take the output.
However, there is one issue: when I run mesh_viz.py, the execution finishes, but nothing is displayed in the output. Also, nothing is saved.
"How is the output?

[Sirui-Xu]

Thanks for the clarification, you may tune the parameter here (

InterDiff/interdiff/eval_smpl_short.py

Line 388 in d584b1f

parser.add_argument("--render_epoch", type=int, default=50)

) and run this evaluation code to get the visualization. Please follow the instructions in the README

[mansooreh1]

Hi, thank you for your responses and the time you’re taking , I need images like Figure 1 in your paper for my own article. Could you please guide me exactly on how to do this? Because I haven’t yet obtained such a visual output.

[Sirui-Xu]

Hello, we utilize Blender for rendering our outputs. For more details, please refer to the visualization section of our codebase on GitHub: https://github.com/neu-vi/HOI-Diff/blob/main/README.md#visualization. Essentially, we concatenate the generated results and then go with the same Blender pipeline to produce the teaser.

[mansooreh1]

Hello, thank you for your helpful guidance.
When I run python -m visualize.render_mesh --input_path /path/to/mp4/stick/figure/file, it throws an error saying that the results.npy file does not exist.
If I’m not mistaken, the input_path should contain multiple frames of a human-object interaction that are output as an MP4 file.
However, I don't understand what results.npy is and how it is supposed to be generated.

[Sirui-Xu]

Hey, sorry for the late reply! I think you can save your generated results as .npy files containing the vertices and faces of the human and object. Then, you can use the command to load the .npy files and visualize them with Blender.

[mansooreh1]

Thank you very much for your reply

[mansooreh1]

Hello. Despite all my efforts, I was not able to perform the visualization. I couldn’t generate the results.npy file and the mp4 files. Could you please explain the steps to produce the results.npy file and the mp4 files in more detail?I am a beginner and would be very grateful if you could provide a detailed explanation for me. I was unable to produce the npy files.