Cervical cancer is the fourth most common cancer among women worldwide, according to the World Health Organization (WHO). Prevention and screening through the Papanicolaou test are essential strategies, but they significantly increase the workload of cytopathologists. In this work, we present a new public dataset of liquid-based cytology images, containing squamous cells annotated according to the Bethesda system. Created from real samples, the dataset features detailed bounding box annotations that fully enclose individual cells–an uncommon characteristic among existing public datasets. To assess its applicability, we propose a deep learning model that combines YOLO for detection and EfficientNet for classification into 6, 3, and 2 classes. To evaluate generalization and enable comparison with prior work, we also included the publicly available CRIC dataset in the experiments. The model achieved detection scores of mAP@0.5 = 0.926 and mAP@0.5:0.95 = 0.669. For the classification tasks, it obtained recall values of 0.61, 0.83, and 0.96, respectively, reaching or exceeding the typical sensitivity range reported for human observation in some scenarios. These results highlight the relevance of the proposed dataset as a contribution to the scientific community, the model’s ability to generalize across different cytology settings, and the overall potential of this work to assist medical professionals in cervical cancer screening.