AD is a neurodegenerative condition that affects brain cells. It is a progressive and incurable disease. Early detection will save the patient’s brain cells from further damage and thereby prevent permanent memory loss. Various automated methods and procedures for the detection of Alzheimer’s disease have been proposed in recent years. To mitigate the loss to a patient’s mental health, many approaches concentrate on quick, reliable, as well as early disease detection. Even though deep learning techniques have greatly enhanced imaging devices for medical purposes for Alzheimer’s disease by delivering diagnostic accuracy that is like that of humans. The existence of strongly associated features in the brain structure continues to be a challenge for multi-class classification. This used a ResD hybrid approach based on Resnet18 and Densenet121 for the multiclass classification of Alzheimer’s Disease on the MRI dataset. Information from the two pre-trained models is combined for classification. Experiments show that the proposed hybrid model outperforms alternative techniques from existing works. The proposed ResD model gives a weighted average (macro) precision score of 99.61%. Through experiments, we show that the proposed hybrid model produces less classification error with hamming loss of 0.003.