The paper introduces SWVR, a lightweight deep learning algorithm for forest fire detection and recognition. SWVR combines the Reparameterization Vision Transformer (RepViT) and Simple Parameter-Free Attention Module (SimAM) to efficiently extract fire-related features with reduced computational complexity. It features a bidirectional fusion network, incorporates lightweight Ghost Shuffle Convolution (GSConv), and uses the Wise Intersection over Union (WIOU) loss function. SWVR achieves 79.6% accuracy in detecting forest fires, a 5.9% improvement over the baseline, and operates at 42.7 frames per second. It reduces the model parameters by 11.8% and computational cost by 36.5%. The paper discusses the effectiveness of each component, including the RepViTBlock, SimAM, VoVGS CSP, and WIOU loss function, and presents experimental results to validate the model's performance. The SWVR model is designed to be lightweight and efficient, making it suitable for real-time forest fire monitoring and early detection.The paper introduces SWVR, a lightweight deep learning algorithm for forest fire detection and recognition. SWVR combines the Reparameterization Vision Transformer (RepViT) and Simple Parameter-Free Attention Module (SimAM) to efficiently extract fire-related features with reduced computational complexity. It features a bidirectional fusion network, incorporates lightweight Ghost Shuffle Convolution (GSConv), and uses the Wise Intersection over Union (WIOU) loss function. SWVR achieves 79.6% accuracy in detecting forest fires, a 5.9% improvement over the baseline, and operates at 42.7 frames per second. It reduces the model parameters by 11.8% and computational cost by 36.5%. The paper discusses the effectiveness of each component, including the RepViTBlock, SimAM, VoVGS CSP, and WIOU loss function, and presents experimental results to validate the model's performance. The SWVR model is designed to be lightweight and efficient, making it suitable for real-time forest fire monitoring and early detection.