The goal of this research project is to compare the accuracy of the CNN method using the Softmax classifier and the OpenCV library by taking out new features and comparing them to the accuracy of the CNN method. To analyse the suggested system, CNN algorithms with softmax classifier and OpenCV library were used with a sample size of 20 to find facial expression recognition by comparing the accuracy and error rate between the techniques. In the proposed study, the dataset was used to collect data on the outcome and accuracy of the OpenCV error rate, and the OpenCV improvement was compared to the CNN softmax classifier technique. The average accuracy of OpenCV over techniques that use convolutional neural networks is 95.5%. The T-Test doesn’t show anything important (p 0.05). Face recognition systems were used to get information from a number of places for the study. Compared to the convolution neural network method, the OpenCV algorithm made the result more accurate and less likely to make mistakes.