Accurate assessment of cell viability is a critical step in biomedical research and diagnostic applications. Conventional methods for distinguishing live and dead cells from microscopic images are usually manual, time-consuming and prone to expert bias. In this study, we explored the application of deep learning-based object detection models such as Yolov5, Yolov7, Yolov8 (with variants s, n, m, l and x) and Faster R-CNN for the automated detection and classification of live and dead cells in microscopic images. A dataset of microscopic cell images was used with each image manually annotated using the LabelImg tool to define bounding boxes around live and dead cells. We evaluated the models using standard detection metrics like precision, mean Average Precision (mAP@50), recall and F1-score. In addition, we also analyzed the computational cost of each model based on specific parameters such as inference time, execution speed (ms/image), number of model parameters and GFLOPs. Yolov8m achieved the best overall performance with precision, mAP@50, recall and F1-score of 0.979, 0.989, 0.958 and 0.968 for live cells and 0.962, 0.962, 0.943 and 0.952 for dead cells. Faster R-CNN demonstrated superior recall for both cell types and high mAP@50, while Yolov5 and Yolov7 models also achieved competitive performance. These results highlight the feasibility of applying modern object detection models for reliable cell viability assessment, significantly reducing manual effort while improving consistency and throughput in biomedical imaging workflows.