Stochastic gradient descent and its derivatives, Adam and RMSprop, are efficient optimization algorithms that speed up the training process, enabling models to converge faster while saving computational resources. Optimization handles complexity by intelligently moving through the parameter space to avoid the trap of saddle spots and local minima. At the same time, optimization is central to regularization techniques that prevent overfitting, hence ensuring good generalization of the models to unseen data. Advanced methods further make it feasible to train models on huge datasets and support adaptive learning rates to improve convergence and stability. Motivated by this, in this research work, the effect of the D-Adaptation algorithm on image classification tasks is assessed. The algorithm was evaluated on a standard image classification model. In this regard, the D-Adaptation technique was leveraged for the optimization of the Adam optimizer, and its performance was evaluated against scenarios without D-Adaptation. Experimental observations assert that the D-Adaptation method is better at improving optimization efficiency and creating better metric results for tasks related to image classification.