The development of cells can be studied by analyzing changes in their shape, movement, and colony size. By photographing cells exposed to specific solutions, these factors can be systematically evaluated. Such studies often generate over two thousand images, requiring experts to manually examine them to derive experimental results. To address this challenge, this research aims to develop an automated method for analyzing images to study cell structural changes. A convolutional neural network (CNN) model was created using transfer learning and trained to classify single cells extracted from colony images through watershed segmentation. The findings of this study indicate that transfer learning can achieve satisfactory precision in classifying highly specific images, with 75% accuracy in training and much higher during manual testing of the model’s accuracy. However, the study results also highlight that such models are heavily dependent on the quality of input data. Consequently, the challenge of clustered cell classification is largely reduced to ensuring precise segmentation and enhancing data quality.