Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Gasformer: A Transformer-based Architecture for Segmenting Methane Emissions from Livestock in Optical Gas Imaging

Toqi Tahamid Sarker, Mohamed G Embaby, Khaled R Ahmed, Amer AbuGhazaleh

TL;DR

Gasformer tackles the challenge of visualizing and quantifying low-flow methane emissions from livestock using Optical Gas Imaging by delivering accurate plume segmentation with a transformer-based encoder and a lightweight decoder. The Mix Vision Transformer encoder extracts multi-scale features, while the Light-Ham decoder with the HamMD module refines segmentation maps efficiently. Two labeled datasets, MR (controlled release) and CR (dairy cow rumen gas), demonstrate Gasformer's superior performance over state-of-the-art baselines, achieving mIoU values of 85.9 and 88.56 respectively, and enabling real-time/near-real-time inference. The work discusses practical limitations of the FLIR GF77 camera, provides thorough ablations to justify architectural choices, and offers datasets and methods to advance methane-emission monitoring and mitigation in agricultural settings.

Abstract

Methane emissions from livestock, particularly cattle, significantly contribute to climate change. Effective methane emission mitigation strategies are crucial as the global population and demand for livestock products increase. We introduce Gasformer, a novel semantic segmentation architecture for detecting low-flow rate methane emissions from livestock, and controlled release experiments using optical gas imaging. We present two unique datasets captured with a FLIR GF77 OGI camera. Gasformer leverages a Mix Vision Transformer encoder and a Light-Ham decoder to generate multi-scale features and refine segmentation maps. Gasformer outperforms other state-of-the-art models on both datasets, demonstrating its effectiveness in detecting and segmenting methane plumes in controlled and real-world scenarios. On the livestock dataset, Gasformer achieves mIoU of 88.56%, surpassing other state-of-the-art models. Materials are available at: github.com/toqitahamid/Gasformer.

Gasformer: A Transformer-based Architecture for Segmenting Methane Emissions from Livestock in Optical Gas Imaging

TL;DR

Gasformer tackles the challenge of visualizing and quantifying low-flow methane emissions from livestock using Optical Gas Imaging by delivering accurate plume segmentation with a transformer-based encoder and a lightweight decoder. The Mix Vision Transformer encoder extracts multi-scale features, while the Light-Ham decoder with the HamMD module refines segmentation maps efficiently. Two labeled datasets, MR (controlled release) and CR (dairy cow rumen gas), demonstrate Gasformer's superior performance over state-of-the-art baselines, achieving mIoU values of 85.9 and 88.56 respectively, and enabling real-time/near-real-time inference. The work discusses practical limitations of the FLIR GF77 camera, provides thorough ablations to justify architectural choices, and offers datasets and methods to advance methane-emission monitoring and mitigation in agricultural settings.

Abstract

Methane emissions from livestock, particularly cattle, significantly contribute to climate change. Effective methane emission mitigation strategies are crucial as the global population and demand for livestock products increase. We introduce Gasformer, a novel semantic segmentation architecture for detecting low-flow rate methane emissions from livestock, and controlled release experiments using optical gas imaging. We present two unique datasets captured with a FLIR GF77 OGI camera. Gasformer leverages a Mix Vision Transformer encoder and a Light-Ham decoder to generate multi-scale features and refine segmentation maps. Gasformer outperforms other state-of-the-art models on both datasets, demonstrating its effectiveness in detecting and segmenting methane plumes in controlled and real-world scenarios. On the livestock dataset, Gasformer achieves mIoU of 88.56%, surpassing other state-of-the-art models. Materials are available at: github.com/toqitahamid/Gasformer.
Paper Structure (20 sections, 1 equation, 5 figures, 8 tables)

This paper contains 20 sections, 1 equation, 5 figures, 8 tables.

Table of Contents

  1. Introduction
  2. Methods
  3. Encoder
  4. Decoder
  5. Dataset
  6. FLIR GF77 Optical Gas Imaging Camera
  7. Controlled Methane Release Dataset
  8. Dairy Cow Rumen Gas Dataset
  9. Mask generation
  10. Experiments and Discussion
  11. Implementation Details
  12. Ablation Study
  13. Results
  14. Segmentation Results on the MR Dataset
  15. Segmentation Results on the CR Dataset
  16. ...and 5 more sections

Figures (5)

  • Figure 1: Overall framework of Gasformer. We extract multi-scale features from the encoders, and the decoder concatenates these multi-level features and predicts the segmentation mask
  • Figure 2: Controlled methane release visualizations at varying flow rates on MR dataset. (a) Flow rate of 10 standard cubic centimeters per minute (SCCM). (b) Flow rate of 60 SCCM. (c) Flow rate of 80 SCCM.
  • Figure 3: (a) CR dataset images. (b) Associated ground truth masks for gas plume regions.
  • Figure 4: Segmentation performance of different models on MR dataset, showing the input images, ground truth masks, and the segmentation outputs for 40 SCCM, 50 SCCM, and 80 SCCM methane flow rates.
  • Figure 5: Semantic segmentation results of CR dataset.