Table 4.
Comparison of deep learning models on their parameters, complexity, strengths, and weaknesses.
| Model | Parameters (MB) | Complexity | Strengths | Weaknesses | |
|---|---|---|---|---|---|
| Total | Trainable | ||||
| Inception—ResNet v2 | 0.29 | 3.01 | Complex (164 layers) | Simple and fast | Low accuracy, overfitting risk |
| Vgg19 | 77.4 | 1.01 | Very Deep (19 layers) | High accuracy | Very deep, computationally expensive |
| ResNet101 | 163.2 | 513.27 | Deep (101 layers) | High generalizability | Requires training time |
| Inception v3 | 87.18 | 4.01 | Complex (48 layers) | Efficient feature Extraction | Overfitting risk |
| Xception | 80.08 | 513.27 | Complex (312 layers) | depth-wise separable convolutions | Computational complexity |
| Inception-ResNet v2 | 0.29 | 3.01 | Complex (164 layers) | Simple and fast | Low accuracy, overfitting risk |
| SSDHR classifier | 11.62 | 11.61 | Highly customizable, | High accuracy | Less computational complexity |