Few-Shot Image Classification — ResNet-50 & EfficientNet (Research/Implementation)
Other
Developed and deployed an image classification few-shot learning pipeline using ResNet-50 and EfficientNet, focusing on transfer learning for low-resource datasets. Applied custom data augmentation strategies such as rotation, cutout, and mixup to reduce overfitting and improve benchmark performance. Conducted model evaluation on 5-way 1-shot and 5-shot classification settings to measure accuracy improvements and baseline comparisons. • Implemented training/evaluation code for few-shot learning experiments • Applied augmentation to existing datasets for generalization • Measured accuracy gains (~12%) versus baseline • Supported research reporting for an IEEE-indexed publication
2025 - 2025