Phase 1: Data Input Occlusion Handling

1. Online Classroom Video Feed

2. Facial Complete Capture [154]

Phase 3: Facial Restoration with Regeneration GAN

4. Input: Occluded Image [197]

Uses a Regenerative Generative Adversarial Network to 'inpaint' [16, 191]

GAN Processing [228]
- Generator Network [193, 232, 233]
- Discriminator Network [194, 231, 238]

6. Output: Reconstructed Image

Phase 4: Emotion Recognition with CNN

Visualization on the Application Dashboard

7. Input: Complete & Facial Classification [224]

Classified emotion data & model metrics visualized [17, 126]

Performance Analytics
- Confusion Matrix (560)
- ROC & Precision Recall (497)
- F1-Score & Accuracy

11. Data Aggregation & Display [110, 182]

Emotion Analytics
- Image Compliance [71]
- Confusion Matrix Metrics [300, 863, 637, 421, 480]
- Real-time Emotion Distribution Heatmaps [127]

12. Educator Interpretation & Action [56, 628]

Final Outcomes
- Adaptive Teaching (20)
- Student Support
- Enhanced Learning Environment

tt