Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

SmokeNet: Efficient Smoke Segmentation Leveraging Multiscale Convolutions and Multiview Attention Mechanisms

Xuesong Liu, Emmett J. Ientilucci

TL;DR

SmokeNet tackles efficient, accurate smoke segmentation in diverse environments by integrating multiscale rectangular convolutions with a multiview linear attention mechanism in a UNet-like encoder–decoder, complemented by a layer-specific loss strategy. The approach achieves strong segmentation performance on synthetic and real-world datasets, including quarry smoke, while maintaining a minimal parameter count and high inference speed. Key contributions include a novel multiscale feature extractor, a lightweight attention module, and a layer-aware loss, along with a quarry smoke dataset to advance generation and evaluation in challenging industrial settings. Practically, SmokeNet enables real-time monitoring and safety management in environments with limited computational resources, contributing to environmental oversight and industrial safety applications.

Abstract

Efficient segmentation of smoke plumes is crucial for environmental monitoring and industrial safety, enabling the detection and mitigation of harmful emissions from activities like quarry blasts and wildfires. Accurate segmentation facilitates environmental impact assessments, timely interventions, and compliance with safety standards. However, existing models often face high computational demands and limited adaptability to diverse smoke appearances, restricting their deployment in resource-constrained environments. To address these issues, we introduce SmokeNet, a novel deep learning architecture that leverages multiscale convolutions and multiview linear attention mechanisms combined with layer-specific loss functions to handle the complex dynamics of diverse smoke plumes, ensuring efficient and accurate segmentation across varied environments. Additionally, we evaluate SmokeNet's performance and versatility using four datasets, including our quarry blast smoke dataset made available to the community. The results demonstrate that SmokeNet maintains a favorable balance between computational efficiency and segmentation accuracy, making it suitable for deployment in environmental monitoring and safety management systems. By contributing a new dataset and offering an efficient segmentation model, SmokeNet advances smoke segmentation capabilities in diverse and challenging environments.

SmokeNet: Efficient Smoke Segmentation Leveraging Multiscale Convolutions and Multiview Attention Mechanisms

TL;DR

SmokeNet tackles efficient, accurate smoke segmentation in diverse environments by integrating multiscale rectangular convolutions with a multiview linear attention mechanism in a UNet-like encoder–decoder, complemented by a layer-specific loss strategy. The approach achieves strong segmentation performance on synthetic and real-world datasets, including quarry smoke, while maintaining a minimal parameter count and high inference speed. Key contributions include a novel multiscale feature extractor, a lightweight attention module, and a layer-aware loss, along with a quarry smoke dataset to advance generation and evaluation in challenging industrial settings. Practically, SmokeNet enables real-time monitoring and safety management in environments with limited computational resources, contributing to environmental oversight and industrial safety applications.

Abstract

Efficient segmentation of smoke plumes is crucial for environmental monitoring and industrial safety, enabling the detection and mitigation of harmful emissions from activities like quarry blasts and wildfires. Accurate segmentation facilitates environmental impact assessments, timely interventions, and compliance with safety standards. However, existing models often face high computational demands and limited adaptability to diverse smoke appearances, restricting their deployment in resource-constrained environments. To address these issues, we introduce SmokeNet, a novel deep learning architecture that leverages multiscale convolutions and multiview linear attention mechanisms combined with layer-specific loss functions to handle the complex dynamics of diverse smoke plumes, ensuring efficient and accurate segmentation across varied environments. Additionally, we evaluate SmokeNet's performance and versatility using four datasets, including our quarry blast smoke dataset made available to the community. The results demonstrate that SmokeNet maintains a favorable balance between computational efficiency and segmentation accuracy, making it suitable for deployment in environmental monitoring and safety management systems. By contributing a new dataset and offering an efficient segmentation model, SmokeNet advances smoke segmentation capabilities in diverse and challenging environments.

Paper Structure

This paper contains 25 sections, 19 equations, 3 figures, 2 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Encoder-Decoder Architectures
  4. Lightweight Segmentation Models
  5. Methodology
  6. Encoder
  7. Multiscale Feature Extraction
  8. Multiview Linear Attention Mechanism
  9. Decoder
  10. Decoder with Skip Connections
  11. Decoder Stage Operations
  12. Loss Function
  13. Experiments
  14. Experimental Setup
  15. Deep Learning Architecture
  16. ...and 10 more sections

Figures (3)

  • Figure 1: Overview of SmokeNet's Architecture. (a) The overall model architecture integrating multiscale convolutions and multiview attention mechanisms. (b) The Multiscale Module capturing spatial information at various scales for accurate smoke segmentation. (c) The Multiview Module enhancing feature refinement through attention mechanisms.
  • Figure 2: Segmentation results of SmokeNet and comparison models on sample images from four test datasets.
  • Figure 3: Segmentation results of SmokeNet in challenging quarry scenarios.