Table 6 Comparison of GRCornShot with other few-shot techniques in the literature.
Ref. | Models used | Datasets | Accuracy (%) | Key features |
|---|---|---|---|---|
Task-driven meta-learning | PlantVillage | Â | Benchmarking meta-learning, analysis of N-way, K-shot | |
PMF+FA with ViT and ResNet50 | PlantDoc | 90.12 | Few-shot learning, Feature Attention Module | |
Faster R-CNN, Siamese network | Strawberry leaf patch dataset | 96.67 | Object detection with Few Shot | |
Inception V3, Siamese network with Triplet loss | PlantVillage | 94.0 | Few-shot learning, reduction in training data | |
Few-shot class-incremental learning | Chrysanthemum dataset | 80.13 | Hyperspectral imaging, addresses unbalanced classes | |
Conditional adversarial autoencoders (CAAE) | Citrus aurantium L. dataset | HarMean 53.4 | Zero- and few-shot recognition | |
ResNet18, ResNet34, ResNet50, DAML | Multiple plant disease datasets | 81–99 | Transfer learning, metric learning, domain shift analysis | |
Frequency domain few-shot learning | Plant disease images in frequency domain | N/A | Discrete cosine transform, frequency selection, Gaussian-like calibration | |
GRCornShot | ResNet 50, Prototypical Network | Kaggle and Roboflow | 97.89 | High accuracy with Few Shot |
