Effective and Precise Detection of Hazardous Fires from CCTV Images Using YOLOv7 Algorithm in Comparison with CNN

To improve the ability to recognize potentially hazardous fires, this research analyzed and contrasted the performance of two different types of deep learning methods: Novel YOLOv7 and Convolutional Neural Networks (CNN). The researchers amassed a dataset consisting of 68 samples, which were divided equally into two groups, each including 34 samples. Group 2 made use of CNN, whereas Group 1 used the Novel YOLOv7 method for their data. The dataset was imported into the Roboflow programme for analysis. The Novel YOLOv7 algorithm was trained on the dataset with the assistance of Google Collab. The number of participants in the experiment was determined with the use of an online statistical programme known as clincalc.com. The pre-test power was set at 80%, and the alpha value was set at 0.05. The size of the sample was established by using the findings of much earlier research. After the investigation was completed, the researchers estimated the mean average accuracy (mAP) for both methods. The mean absolute performance (mAP) of the CNN algorithm was 0.3291, whereas the mean absolute performance (mAP) of the Novel YOLOv7 method was 0.4506. According to these data, there was a statistically significant difference between the two algorithms in terms of accuracy, with a significance value of 0.001 (p0.05). This difference was found to be statistically significant. In conclusion, the Novel YOLOv7 algorithm showed more accurate hazardous fire detection compared to Convolutional Neural Networks (CNN) for the dataset that was presented.