In recent years, the proliferation of online reviews has become pronounced, marking a significant shift in how individuals convey their opinions about various establishments, particularly restaurants. Notably, there has been a discernible trend wherein individuals incorporate images alongside textual feedback, enriching the review experience by providing nuanced visual cues. These images serve as substrates for sentiment analysis, wherein the polarity of sentiments is discerned through either facial expressions or food depiction. While considerable strides have been made in text sentiment analysis, this paper delves into visual analysis as a focal point for discerning sentiment expression in images. Leveraging various deep learning models such as Convolutional Neural Networks (CNNs), VGG19, InceptionV3 and ResNet10 a robust image sentiment prediction model is identified. The primary aim is to efficiently identify sentiment and augment the accuracy of food review datasets disseminated across social media platforms. The findings of this study reveal the performance of each model in sentiment analysis based on a custom dataset compared to various others.