Research Article
[Retracted] 3D Animation Automatic Generation System Design Based on Deep Learning
Table 1
Feature recognition experimental results of different 3D animation facial expression generation methods.
| | Occlusion type | Block body type | Root mean square error | Pearson’s correlation coefficient |
| Expression generation method based on improved CycleGAN model | Unobstructed | 0 × 0 | 0.018 | 0.889 | Random block 25% | 112 × 112 | 0.132 | 0.506 | Random block 50% | 120 × 120 | 0.201 | 0.489 |
| Expression generation method based on generative adversarial network | Unobstructed | 0 × 0 | 0.305 | 0.968 | Random block 25% | 112 × 112 | 0.425 | 0.669 | Random block 50% | 120 × 120 | 0.021 | 0.589 |
| The proposed expression generation method | Unobstructed | 0 × 0 | 0.001 | 0.995 | Random block 25% | 112 × 112 | 0.013 | 0.959 | Random block 50% | 120 × 120 | 0.016 | 0.978 |
|
|