Summary: The paper presents a method for automatically animating children's drawings of the human figure, which is robust, simple, and accessible, and has been implemented in a public website used by millions of people worldwide. The paper also introduces a new annotated dataset of amateur drawings collected via the public demo.

Key insights and lessons learned:

1. Children's drawings of the human figure exhibit inventiveness, creativity, and variance, which pose challenges for animation.
2. The proposed method for animating children's drawings is robust and straightforward, making it accessible to a wide audience.
3. The Animated Drawings Demo, implemented based on the proposed method, has been widely used by millions of people, indicating its broad appeal.
4. The Amateur Drawings Dataset, introduced in the paper, is a valuable resource for further research in the field of animating children's drawings.

Questions for the authors:

1. What are the main technical challenges in animating children's drawings of the human figure, and how does your proposed method address them?
2. Can you provide more details about the experiments conducted to determine the amount of training data needed for fine-tuning?
3. What were the findings of the perceptual study that demonstrated the appeal of the twisted perspective retargeting technique?
4. How do you envision the use of the Amateur Drawings Dataset in future research, and what potential applications do you see for it beyond the scope of this paper?
5. Can you discuss any limitations of your proposed method and potential directions for future improvements?

Suggestions for related topics or future research directions:

1. Exploring the use of machine learning techniques for animating other types of artistic or creative expressions, such as paintings or sculptures.
2. Investigating the potential applications of the proposed method in educational settings, such as using it as a tool for teaching children about human anatomy and movement.
3. Examining the impact of cultural and developmental factors on children's drawings and how they may affect the animation process.
4. Investigating the potential use of the proposed method in fields beyond art and entertainment, such as in healthcare for rehabilitation exercises or in virtual reality for avatar customization.
5. Exploring the ethical considerations of using machine learning techniques for animating children's drawings, including issues related to privacy, consent, and representation.

Relevant references:

1. Li, Y., Liu, L., Li, X., Xu, Y.Q., Yang, C., & Tai, C.L. (2018). A closed-form solution to retargeting 3D human body shapes. ACM Transactions on Graphics (TOG), 37(6), 1-14.