Aim: The aim of the project is to improve the classification accuracy in down syndrome detection using ResNet34 in comparison with ResNet18 classifiers. Materials and Methods: The dataset of the ResNet34 and ResNet18 is taken from Kaggle open access data. A total of 20 samples are collected from a dataset from Kaggle. Total 20 samples are used for two groups and 10 samples are used ResNet34 (group-1) and it is compared with ResNet18 (group-2). The total samples were calculated using sample computation with pretest power of 80 % where Alpha is 0.05 and Beta is 0.2. Result: From MATLAB simulation, ResNet34 achieved an accuracy rate of 87.4 % compared to accuracy rate of 82.5 % by ResNet18. The statistical analysis was calculated and done by performing an independent variable t t-test with significance value of 0.002 (p<0.05). Conclusion: It is concluded that the entropy, energy, contrast and brightness feature extraction show higher accuracy of the ResNet34 algorithm compared to the ResNet18 algorithm from the considered dataset.