Table 1.
Comparative review of related works.
| Study | Year | Architecture | Used Dataset | Val_accuracy |
|---|---|---|---|---|
| [17] | 2021 | DCNN | FER2013, SFEW2.0, and RAF-DB | 76.69 %, 58.68 %, and 87.13 %, respectively |
| [13] | 2022 | CNN based on VGGNet | FER2013, FER+ | 79 % |
| [18] | 2022 | 3 stage CNN | FER2013 | 82 % |
| [14] | 2023 | ERCNN | FER2013, FER+ | 82.64 % |
| [15] | 2023 | EfficientNet-XGBoost | CK+ and FER2013 | 100, 72.5 % respectively |
| [10] | 2023 | FER-CHC | FER2013 | 74.68 % |
| [16] | 2023 | Proposed + ResNet-50 | FER-2013, CK+, RAF-DB | 73 %, 89 % and 76 %, respectively. |
| [19] | 2023 | SSF-ViT (L) | FER2013 | 74.95 % |
| [20] | 2023 | CNN-based Inception-v3 architecture | FER2013 | 73.09 % |
| [21] | 2023 | Xception Net | FER2013 | 77.92 % |
| [22] | 2023 | EmoNAS | FER2013 | 67.9 % |
| [23] | 2023 | SSA-NET | FER2013 | 67.57 % |
| [24] | 2024 | EduViT based on the MobileViT architecture | FER2013 | 66.51 % |
| [25] | 2024 | Hybridized CNN-LSTM | FER2013 | 79.34 % |
| [26] | 2024 | Activation-matrix Triplet loss and pseudo label with Complementary Information | FER2013 | 71.62 % |
| [27] | 2024 | EfficientNet | FER2013 | 58.41 % |
| [28] | 2024 | Custom CNN | FER2013 | 57.4 % |