This paper provides an attempt to classify the food images using the CNN and with the support of EfficientNet-B1 using Food-101 database. The classification of food is an important task in different domains such as dietary assessment and automatic food intake recording. The dataset which is publicly available is the Food-101 dataset containing 10,100 images in 101 classes. We used baseline CNN and efficient CNN called EfficientNet-B1 because of its better performance compared to the number of parameters. Specific standard approaches, including data augmentation and transfer learning, were applied for a higher accuracy of models participating in the evaluation. We demonstrated throughout our experiments that EfficientNet-B1 is superior to the CNN baseline model in both accuracy and the time taken during training and inference. In this work, we demonstrate the applicability of EfficientNet-based architectures for large scale food classification problems, which should be helpful for future real-world large scale food recognition systems.