One-shot and few/low-shot machine learning are novel techniques that use less data in sequence learning for predictive analysis. These techniques have been applied to image databases to further segment data and create forecasting models. In this paper, a financial dataset is converted and built into an image database of five feature classes. One-shot and few-shot learning models, using prototypical networks and matching networks, were tested on the constructed financial image database to forecast foreign exchange (Forex) rates, specifically comparing the Euro against the US Dollar (EUR/USD). A comparison study was also conducted using a meta-learner Long Short-Term Memory (LSTM) model to forecast the same exchange rate. Besides, the tuning of hyperparameters for both few-shot learning and LSTM models were examined. LSTM results were compared with both one-shot and few-shot learning models to evaluate the effectiveness of each model, indicating the possibility of achieving a better performance using few-shot learning techniques on forex dataset.