The aim of this study is to develop and evaluate a deep learning deep neural network model to accurately identify and classify bacteria using DIBAS datasets. We use Efficient architecture, a lightweight and efficient convolutional neural network (CNN) to process image data and extract features. Through a carefully designed experimental flow, our model is extensively trained and validated on DIBAS datasets. The experimental results show that our deep neural network model has achieved a success rate of more than 80% in the bacterial recognition task, demonstrating high accuracy and robustness. In addition, we compare the effects of different hyperparameter settings on model performance and explore the model’s performance when dealing with unbalanced data sets. Our results not only provide a new and efficient tool for the field of bacterial identification, but also provide a valuable reference for future research in similar fields.