Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Promptable Fire Segmentation: Unleashing SAM2's Potential for Real-Time Mobile Deployment with Strategic Bounding Box Guidance

Emmanuel U. Ugwu, Zhang Xinming

TL;DR

This work addresses real-time fire segmentation by systematically evaluating SAM2.1 variants and mobile PROMPTable configurations under bounding-box prompting. It demonstrates that bounding-box prompts, especially Box+MP, substantially enhance segmentation quality on fire datasets, with lightweight variants like TinySAM and MobileSAM offering favorable efficiency for edge scenarios albeit without achieving true real-time performance. Across three datasets, the study benchmarks accuracy versus efficiency, highlighting the gap between zero-shot SAM2.1 capabilities and supervised domain-specific methods but underscoring the potential for practical deployment via prompt design and hardware-aware models. The findings establish benchmarks and practical guidelines for deploying promptable segmentation in fire monitoring, informing future work toward real-time, resource-constrained surveillance systems.

Abstract

Fire segmentation remains a critical challenge in computer vision due to flames' irregular boundaries, translucent edges, and highly variable intensities. While the Segment Anything Models (SAM and SAM2) have demonstrated impressive cross-domain generalization capabilities, their effectiveness in fire segmentation -- particularly under mobile deployment constraints -- remains largely unexplored. This paper presents the first comprehensive evaluation of SAM2 variants for fire segmentation, focusing on bounding box prompting strategies to enhance deployment feasibility. We systematically evaluate four SAM2.1 variants (tiny, small, base_plus, large) alongside mobile-oriented variants (TinySAM, MobileSAM) across three fire datasets using multiple prompting strategies: automatic, single positive point (SP), single positive point + single negative point (SP+SN), multiple positive points (MP), bounding box (Box), and hybrid variants (Box+SP and Box+MP). Our experimental results demonstrate that bounding box prompts consistently outperform automatic and single point-based approaches, with Box+MP achieving the highest mean IoU (0.64) and Dice coefficient (0.75) on the Khan dataset. Lightweight variants such as TinySAM and MobileSAM further reduce memory and computational costs, making them more suitable for latency-tolerant edge scenarios. Overall, this work provides critical insights for deploying promptable segmentation models in fire monitoring systems and establishes benchmarks for future research in domain-specific SAM applications. Code is available at: https://github.com/UEmmanuel5/ProFSAM

Promptable Fire Segmentation: Unleashing SAM2's Potential for Real-Time Mobile Deployment with Strategic Bounding Box Guidance

TL;DR

This work addresses real-time fire segmentation by systematically evaluating SAM2.1 variants and mobile PROMPTable configurations under bounding-box prompting. It demonstrates that bounding-box prompts, especially Box+MP, substantially enhance segmentation quality on fire datasets, with lightweight variants like TinySAM and MobileSAM offering favorable efficiency for edge scenarios albeit without achieving true real-time performance. Across three datasets, the study benchmarks accuracy versus efficiency, highlighting the gap between zero-shot SAM2.1 capabilities and supervised domain-specific methods but underscoring the potential for practical deployment via prompt design and hardware-aware models. The findings establish benchmarks and practical guidelines for deploying promptable segmentation in fire monitoring, informing future work toward real-time, resource-constrained surveillance systems.

Abstract

Fire segmentation remains a critical challenge in computer vision due to flames' irregular boundaries, translucent edges, and highly variable intensities. While the Segment Anything Models (SAM and SAM2) have demonstrated impressive cross-domain generalization capabilities, their effectiveness in fire segmentation -- particularly under mobile deployment constraints -- remains largely unexplored. This paper presents the first comprehensive evaluation of SAM2 variants for fire segmentation, focusing on bounding box prompting strategies to enhance deployment feasibility. We systematically evaluate four SAM2.1 variants (tiny, small, base_plus, large) alongside mobile-oriented variants (TinySAM, MobileSAM) across three fire datasets using multiple prompting strategies: automatic, single positive point (SP), single positive point + single negative point (SP+SN), multiple positive points (MP), bounding box (Box), and hybrid variants (Box+SP and Box+MP). Our experimental results demonstrate that bounding box prompts consistently outperform automatic and single point-based approaches, with Box+MP achieving the highest mean IoU (0.64) and Dice coefficient (0.75) on the Khan dataset. Lightweight variants such as TinySAM and MobileSAM further reduce memory and computational costs, making them more suitable for latency-tolerant edge scenarios. Overall, this work provides critical insights for deploying promptable segmentation models in fire monitoring systems and establishes benchmarks for future research in domain-specific SAM applications. Code is available at: https://github.com/UEmmanuel5/ProFSAM
Paper Structure (14 sections, 5 equations, 3 figures, 4 tables)

This paper contains 14 sections, 5 equations, 3 figures, 4 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Methodology
  4. Prompting Strategy Design
  5. Experimental Setup
  6. Datasets
  7. Evaluation Metrics
  8. Implementation Details
  9. Results and Discussion
  10. Overall Performance Analysis
  11. Cross-Dataset Generalization
  12. Video Segmentation and Computational Efficiency
  13. Discussion
  14. Conclusion

Figures (3)

  • Figure 1: Fire segmentation pipeline using SAM2. An input fire video frame is processed with bounding box prompts through the SAM2 model, producing a temporally consistent segmentation mask. The output is visualized as a green fire overlay on the original frame for real-time monitoring applications.
  • Figure 2: Step-by-step fire segmentation using SAM2. From left to right: (1) original fire image, (2) ground truth mask, (3) fire region detection using YOLOv11, (4) SAM2-generated mask using the YOLOv11 bounding box as prompt, and (5) final fire segmentation visualized as a red overlay on the original image.
  • Figure 3: Comparison of computational efficiency metrics for all fire videos using bounding box-based segmentation. Metrics include (a) model size, (b) average inference time, (c) inference speed (FPS), and (d) peak memory usage for different models.