Table 1 Summary of the most recent related work.
Refs. | Year | Dataset used | Dataset size | Proposed methods | Predicted classes | Accuracy |
|---|---|---|---|---|---|---|
2024 | Kaggle ASD Dataset | 2940 images | ResNet34, ResNet50, VGG16, VGG19, AlexNet, MobileNetV2 | ASD vs Non-ASD | ResNet50: 92%, VGG19: 87% | |
2023 | Kaggle ASD Dataset | 3014 images | VGG16, VGG19, EfficientNetB0 | ASD vs Non-ASD | VGG16: 84.66%, VGG19: 80.05%, EfficientNetB0: 87.9% | |
2023 | Kaggle ASD Dataset | 3014 images | MobileNetV2, ResNet50V2, Xception | ASD vs Non-ASD | Xception: 98.9%, ResNet50V2: 97.1%, MobileNetV2: 91.4% | |
2023 | Kaggle ASD Dataset | 2936 images | VGG16, VGG19 | ASD vs Non-ASD | VGG16: 86.33%, VGG19: 84.00% | |
2023 | Private Dataset | 125 toddlers | Multimodal Machine Learning System (MMLS) based on response to name (RTN) | ASD vs Non-ASD | Computer-rated: 74.8%, Human-rated: 82.9% | |
2023 | Custom Video Dataset | 105 children (ASD: 62, Non-ASD: 43) | CNN-based system using facial attributes (expressions, AUs, arousal, valence) | ASD vs Non-ASD | F1 Score: 76%, Sensitivity: 76%, Specificity: 69% | |
2023 | ASD Children Dataset | 2926 images | Vision Transformer (ViT), Knowledge Distillation (ViTASD) | ASD vs Non-ASD | Accuracy: 94.50% | |
2023 | Kaggle ASD Dataset | 2540 images | MobileNet, Xception, Inception V3, EfficientNetB0, EfficientNetB7, and VGG16 | autistic and non-autistic | Accuracy: 88%, 87.7%, 86.1%, 85.6%, 82.6%, and 86.3% respectively | |
2022 | Kaggle ASD Dataset (Version 9) | 2936 images | AutoML (TPOT), Traditional ML, CNN | ASD vs Non-ASD | AutoML: 96%, CNN: 89% | |
2022 | Kaggle ASD Dataset | 2940 images | Xception, VGG19, NASNetMobile | ASD vs Non-ASD | Xception: 91%, VGG19: 80%, NASNetMobile: 78% | |
2022 | Private Dataset | 120 participants | Ambient Facial Image Grouping Task (Oddball Detection) | Autistic vs Non-Autistic | Accuracy not explicitly stated, but results show a significant difference between autistic (65.96%) and non-autistic participants (74.71%) | |
2022 | Kaggle ASD Dataset | 3014 images | MobileNet, Xception | ASD vs Non-ASD | MobileNet: 95%, Xception: 94% | |
2021 | Kaggle ASD Dataset | 3,014 images | MobileNet | ASD vs Non-ASD | 94.64% | |
2021 | Private Dataset (Elim Autism Rehabilitation Center) | 1122 ASD images, 561 TD images | VGG16 | ASD vs Non-ASD | Accuracy: 95%, F1 score: 0.95 |
