Advanced IoT-Based Fire and Smoke Detection System leveraging Deep Learning and TinyML

This research paper presents the development of an advanced IoT-based fire and smoke detection system leveraging deep learning and TinyML techniques. The objective is to create a real-time detection system optimized for deployment on resource-constrained devices like Raspberry Pi. The proposed system aims to provide accurate and timely detection of fire and smoke in indoor environments using computer vision and machine learning.To achieve this, state-of-the-art deep learning models, including MobileNet, EfficientNet, ResNet, InceptionNet, and XceptionNet, are employed, as they have demonstrated promising results in image recognition tasks. These models are optimized for real-time operation and adapted to run efficiently on low-power devices like Raspberry Pi. Techniques such as quantization, pruning, and other TinyML-specific optimizations are utilized to reduce computational cost and memory usage while maintaining high levels of accuracy.The performance of the proposed system is evaluated through extensive testing in various indoor environments. The evaluation focuses on accuracy, response time, and computational efficiency, comparing the system against existing fire and smoke detection systems. The results demonstrate the feasibility and effectiveness of deploying state-of-the-art deep learning models on resource-constrained devices, highlighting the system's ability to provide reliable and cost-effective fire and smoke detection.