Visual Sentiment Classification on DeepSent three-agree ⟨σ3, P2⟩
88AccuracyMLLMsent
Evaluation Results
| Method | Links | |
|---|---|---|
| MLLMsentTraining dataset=PerceptSent, Fine-tuned on DeepSent=false, Evaluation mode=Cross-dataset generalization2025.08 | 88 | |
| ResNet50Architecture=Deep CNN2025.08 | 81.5 | |
| DenseNet169Architecture=Deep CNN2025.08 | 81.4 | |
| InceptionV3Architecture=Deep CNN2025.08 | 79.6 | |
| VGG16Architecture=Deep CNN2025.08 | 76.5 | |
| Campos et al.2025.08 | 74.9 | |
| You et al.2025.08 | 68.7 | |
| LCHType=Hand-crafted2025.08 | 66.4 | |
| GCHType=Hand-crafted2025.08 | 66 |