search

facebookresearch / banmo

BANMo Building Animatable 3D Neural Models from Many Casual Videos
Other
530 stars 59 forks source link

The problem of extracting masks for deformable dynamic target objects #80

Open litchisea opened 4 months ago

litchisea commented 4 months ago

Hello! I want to use Segment Anything Model to extract masks for cats or other animals. How should I change the code? I made some attempts myself and used SAM to extract the masks after video framing, but the masks I extracted were black and white. I don’t know how to add them to the data preprocessing of the project, or how to change the code. In the visualization results, the background of the mask extracted by your data preprocessing is black and the foreground is blue. Why is it blue? Do I need to change my foreground to blue to achieve this? Looking forward to your reply, thank you very much! 00033 00032

gengshan-y commented 4 months ago

Hi, you might want to look at this file that computes masks, and these lines assigns color to the foreground.

litchisea commented 4 months ago

Thank you very much for your prompt reply! The issue has been resolved. May I also ask how the specific ablation experiment was conducted and which part of the code needs to be modified to achieve two ablation experiments?

gengshan-y commented 1 month ago

Hi, could you explain which ablation were you referring to?