Early detection and monitoring of drug impact on cancerous cells are essential for treatment and livelihood. Traditional methods of drug impact detection are often labor-intensive and time-consuming. Although several algorithms have shown their utility in detecting activity inside the cell using biospeckles images. To incorporate automated monitoring and classification, convolutional neural networks (CNNs) are getting attention for analyzing biospeckles. In this paper we deployed CNN architectures: AlexNet, MobileNet, and VGG16 on biosepckle image dataset. A comprehensive dataset of biosepckles images of melanoma cells, categorized into cells with drug-induced and without drug, was used for the model implementation and evaluation. Images were preprocessed and augmented for model generalization. Each model was trained individually. Their performance is compared against the custom dataset generated by data augmentation.