Precise classification and segmentation of Chlorella vulgaris cells are crucial for their use in biotechnology, aquaculture, biofuel production, and environmental monitoring. Manual counting via traditional methods is not only time-consuming but also error-prone, thereby necessitating automatic methods. The work investigates the utilization of EfficientNetV2-S and InceptionResNetV2 for automatic detection, classification, and segmentation of Chlorella vulgaris cells from the CIDACC dataset. The dataset provides 628 high-resolution microscopic images annotated with bounding boxes and segmentation masks. The models were trained on the dataset, with EfficientNetV2-S with 82.65% accuracy, surpassing InceptionResNetV2 (65%), and ResNet50 (73%) in classification. However, InceptionResNetV2 performed better in segmentation, with a mean IoU of 0.884. Grad-CAM visualization was used for better model interpretability and revealed that EfficientNetV2-S focused on cell borders and individual structures, whereas InceptionResNetV2 focused on general cluster patterns. Despite high performance, factors such as overlapping cell clusters and low-contrast images affected model accuracy. This work stresses the need for choice of model architectures according to application needs EfficientNetV2-excels as accurate cell counting while InceptionResNetV2 excels at segmentation and morphological evaluation. Self-supervised learning, attention, and practical validation should be the focus of future work to drive robust models. The results are promising that deep learning-based cell analysis has the potential to significantly improve automated microalgae research with scalable solutions for biotechnological and environmental applications.