Summary
- Jilin Animation Institute built a practical teaching path integrating generative AI with a Transformer-based human pose estimation model for animation majors.
- A controlled comparison found the AI-integrated group outperformed traditional instruction on both technical application and artistic creation measures.
- The optimised pose-estimation model reduced error to 0.15 and lifted animation-quality scores to 92 points on the study's scale.
Context
Empirical, animation-specific evidence on what AI-integrated teaching actually produces is scarce; most studies measure attitudes or general creativity rather than animation skills. This study set out to measure outcomes directly with animation-major students.
What they did
The teaching path combined generative AI tools with a Transformer-based human pose estimation model optimised for animation tasks, embedded into the practical instruction for animation majors.
How it was implemented and tested
The study ran a controlled comparison between an AI-integrated teaching group and a traditional-instruction group, measuring both technical application and artistic creation outcomes rather than perceptions alone.
What they found
The AI-integrated group significantly outperformed the traditional group on both measures. The optimised pose-estimation model reduced error to 0.15 and raised animation-generation quality scores to 92 points. The authors note the Jilin (China) context; some assessment assumptions may not transfer directly.
Materials
Source: Discover Artificial Intelligence (Springer Nature) ↗ (curated external)