This paper presents an optimized model that uses an optimized CNN to detect depressive symptoms from image posts. This is with a view to detecting depression symptoms in individuals. Visual data were collected in their raw form and assessed as having or not having a mental condition. The images were processed, and the relevant features retrieved from them. An optimized convolutional neural network (CNN) was used to simulate the defined classification model of the image posts. The model was implemented using Python Programming Language. Precision, recall, accuracy, and the area under the Receiver Operating Characteristics (ROC) curve were used as performance indicators to assess the model's efficacy. The collected findings indicate that 77% accuracy is achieved by the optimized model. As a result, 77% of the cases were accurately predicted by the model, suggesting that the model is generally accurate in its predictions. The research will contribute to a decrease in the incidence, prevalence, and recurrence of mental health illnesses as well as the disabilities they cause.