FTA-DETR: An efficient and precise fire detection framework based on an end-to-end architecture applicable to embedded platforms

Timely fire alarms are crucial as they can save lives and avoid major economic losses. However, due to the complexity of the structure, the current mainstream DETR-based fire detection models are problematic in terms of practicality because they require large amounts of memory and long inference times. Meanwhile, high-quality fire detection datasets are very scarce, severely limiting the performance of the algorithms. To address these challenges and improve accuracy in complex fire environments, first, we introduce a dataset quality enhancement framework based on diffusion modele (DDPM) to improve the quality of low-quality fire alarm datasets. Second, we propose a novel Deformable-DETR-based fire detection framework (FTA-DETR). Among the innovative optimizations of FTA-DETR, first, we introduce a trainable matrix in the encoder to compute features, which reduces the computational burden of the encoder, highlights compelling features, and significantly reduces the training time. Second, we improve the encoding block by alternately updating high-level and low-level features, greatly reducing the amount of feature computation required for effective detection. This encoder structure is compatible with any state-of-the-art transformer decoder. Next, to accommodate the multi-scale nature of fires and different environmental complexities, we modify the loss function to WiouV3, which not only speeds up the convergence of the model but also improves the performance. Finally, we smoothly combine FTA-DETR with an acceleration engine like TensorRT to improve inference speed with little loss of accuracy. Compared to other deep learning methods, our approach greatly simplifies the detection process and establishes an end-to-end detection framework. Experiments demonstrate the high quality of datasets generated by the diffusion model-based dataset quality enhancement framework, and these datasets greatly improve the mAP of FTA-DETR (by 2.42% overall). In addition, FTA-DETR outperforms almost all fire detection frameworks in terms of detection accuracy and immunity to interference due to all YOLO-series and DETR-series algorithms in terms of inference speed on a high-performance graphics card A100, with an FPS as fast as 211, and runs stably on small embedded platforms like the Jetson Orin Nano, which has very little computational power, with FPS reaches 76. Overall, the algorithm proposed in this paper outperforms existing mainstream fire detection algorithms across various performance metrics. This superiority is clearly demonstrated in the test results on public datasets: our algorithm achieved an accuracy of 98.32% on the Mivia dataset, a recall rate of 98% on the BoWFire dataset, and an impressive precision of 99.55% on the FireNet dataset. The code is available at https://github.com/wanggoat/FTA-detr.