HI ? If I have two classes, one containing 22 images and the other containing 35 images, is few-shot learning suitable for this case or its better to use Data Augmentation + Transfer Learning? THANKS
Hi @AMILIA
Few-shot learning is designed for scenarios where there are very few examples per class, often in the range of 1 to 10 images per class. Your case is still a low-data scenario but not necessarily in the extreme few-shot range. Using data augmentation to artificially increase your dataset size and transfer learning with a pre-trained model would likely be more effective. However, if the dataset remains too small even after augmentation, few-shot learning techniques could still be explored.
Hope it helps! Feel free to ask if you need further assistance.
1 Like