It is essential for traders to accurately predict changes in the Forex markets in order to make better decisions in dynamic financial markets. This research utilizes the concept of the Fair Value Gap (FVG), which represents the price imbalance between buyers and sellers. To detect patterns, FVG offers key insights into a market reversal or continuation. However, manual detection is error-prone and inefficient in a volatile market. This study proposes an automatic framework that uses convolutional neural networks (CNNs) to efficiently analyse candlestick images and identify FVG patterns. The model is trained on a large number of historical market data sets. The results highlight the potential of CNNs to improve algorithmic trading strategies. For real time validation, our future enhancements will integrate the model into high frequency trading platforms. In addition, we will explore the transformer architecture to improve the detection of complex price movements.