The ambition of this research work is to evaluate the fir and smoke identification in current strategy is by performing a comparative analysis between Convolutional Neural Network (CNN) and MobileNet algorithms. The moto is to increase the accuracy of fire and smoke identification by exploring CNN and MobileNet, and the dataset contains a sample size (N) of 10 instances. The focus is on evaluating key performance metrics such as accuracy, mean, and standard deviation of CNN and MobileNet. Based on a deep learning approach using CNN and Mobilenet, the real-time fire and smoke detection in surveillance systems were studied. The dataset is divided into two groups (one for each algorithm) and the analysis is based on a sample size of 10 instances per category. The objective is to examine the evaluation of CNN and MobileNet in terms of mean, accuracy, and standard deviation. In the comparison of CNN and MobileNet for the identification of smoke and fire, the CNN outperforms an accuracy of 90.6% and MobileNet archives 85.8%.