This research paper suggests a classification methodology of Sickle Cell Disease (SCD) using a deep learning and transfer learning approach. Precisely, the transfer learning models that have been used are MobileNet and VGG-19 to conduct classification tasks on a Sickle Cell image dataset for training and evaluation. Image preprocessing and data augmentation measures were employed to enhance the quality of the datasets and boost the performance of the models, which enhanced variability in the data and robustness in the models. Also, Support Vector Machine (SVM) and Random Forest classifiers were ablation tested to analyse the effect of each individual model component and optimise performance by hyperparameter tuning. Detailed statistical analysis has been conducted, and adversarial attacks have been used to test model stability on perturbed conditions. According to the findings, the VGG-19 model with a SVM classifier was better than the other methods in regards to accuracy. Moreover, this model proved to be very precise, recalls, and F1-score, especially when applied to a smaller part of the dataset, which proves the efficiency it has in detecting normal and sickle-like cells.