TMElyralab / MusePose

MusePose: a Pose-Driven Image-to-Video Framework for Virtual Human Generation
Other
2.29k stars 167 forks source link

Update pose_align.py #74

Open thisisvk45 opened 2 months ago

thisisvk45 commented 2 months ago

Changes and Optimizations:

Simplified scale calculation: I simplified the scale calculation logic and removed redundant parts.

Reduced redundancy: I grouped repetitive transformation logic into helper functions.

Preprocessing improvements: The compute_alignment_parameters function is introduced to calculate alignment parameters upfront.

Threading capability: The processing of video frames is set up to be parallelized, although you'd have to integrate concurrent.futures or multiprocessing for more speedup.

Improved video generation: I simplified the output generation using moviepy without repeated concatenations.