NTU Computer Graphic 2023 Final Project

Description

We re-implemented and modified the paper "A Method for Animating Children's Drawings of the Human Figure" (SIGGRAPH 2023) as our final project. This project is written in python, feel free to fork and play with yourself!
For more, please check out our demo slide: link

Member

we are from NTU GICE DISP Lab (台大電信所R11)

• 顏子鈞
• 許宸睿
• 李振勳

Dependencies

please install them using pip install -r requirements.txt.

Execution

To obtain the motion-retargetted video of a selected drawing figure:

$ python main.py --drawing <figure_name> --motion <motion_num> --strip <strip_num> --sk_pts <sk_pts>

where:

• <figure_name> can be anyone of figures located in config.py. Default = stickman1.
• <motion_num> is the name of target motion video located in target_motion_data/. For example, inputting <motion_num> as 14 will choose the video 14.mp4
• <strip> is the triangle mesh resolution, you can choose 2 or 4. (Note: smaller value requires larger computation)
• <sk_pts> is the total skeleton point used in motion retargetting. Default = 19. You can choose 15/17/19.

For more detail of args, please see main.py.

To stitch the motion-retargetted video and background video together:

$ python modify_video_background.py --pos <leftOrRight> --vid2 <background_video_path> --output <output_path>

where:

• <pos> is left or right. This is where you stitch your video.
• <vid2> is the path to background video.
• <output_path> is the path of output video. Default = output/test_background_output.mp4

For more detail of args, please see modify_video_background.py.

Example Result

drawing	motion retargeted video

References

This implementation is based on:

• "A Method for Animating Children's Drawings of the Human Figure" (SIGGRAPH 2023)
• "Implementing As-Rigid-As-Possible Shape Manipulation and Surface Flattening" (Journal of Graphics Gpu & Game Tools, 2009)
• facebookresearch/AnimatedDrawings
  • we only use code in animated_drawings/model/arap.py: see here link