This model surpasses the performance of the model presented in the DeepFood Paper, both trained on the same dataset. I employed mixed precision training and EfficientNetB0, optimizing the training process by implementing efficient data pipelines using tf.data API methods like batch() and prefetch().

DeepFood achieved an accuracy of 77.4%, whereas this model achieved an even higher accuracy of 80%. Notably, the training time for DeepFood's model spanned 2-3 days, while this model was trained in a remarkably shorter duration—just 30 minutes.