S
SARA·Detect
Publications

Peer-reviewed output from the grant

Three publications trace the system from single-gas detection to a unified, diagnosis-ready framework.

2
published
1
under review
01
ISVC 2025 SPRINGER PUBLISHED

CarboFormer: A Lightweight Semantic Segmentation Architecture for CO₂ Detection Using Optical Gas Imaging

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

International Symposium on Visual Computing (ISVC 2025) · Lecture Notes in Computer Science, Springer · DOI 10.1007/978-3-032-14495-9_1

Introduces a compact transformer-based segmentation network tailored to optical gas imaging. Pre-trained on controlled CO₂ release and fine-tuned on real cattle recordings, CarboFormer reaches 92.98% mIoU with just 5.07M parameters and runs at 84.68 FPS — making real-time, on-device CO₂ plume detection practical for livestock monitoring.

Springer ↗ arXiv:2506.05360 ↗
02
WACV 2026 IEEE/CVF · WORKSHOP PUBLISHED

FUME: Fused Unified Multi-Gas Emission Network for Livestock Rumen Acidosis Detection

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

Proc. IEEE/CVF Winter Conf. on Applications of Computer Vision (WACV) Workshops, 2026 · HARVEST-Vision · pp. 510–519

Extends single-gas detection to a unified, multi-gas setting. FUME fuses representations to jointly segment CO₂ and methane plumes from livestock rumen emissions in a single network — capturing the combined gas signature that better reflects fermentation activity and animal health.

CVF Open Access ↗ PDF ↗
03
UNDER REVIEW SMART AGRICULTURAL TECHNOLOGY

VLMDual: A Vision-Language Distillation Framework for Joint Classification and Plume Segmentation in Rumen Acidosis Detection

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

Manuscript under review · Smart Agricultural Technology (Elsevier)

Closes the loop from perception to diagnosis. VLMDual distills knowledge from large vision-language models into an efficient network that simultaneously classifies acidosis risk and segments gas plumes — pairing what the model sees with the judgment it needs to support on-farm decisions.

In peer review — citation forthcoming
Funding & Acknowledgment

This work is supported by the United States Department of Agriculture, National Institute of Food and Agriculture (USDA-NIFA), through the Capacity Building Grants for Non-Land-Grant Colleges of Agriculture.

GRANT NO.2023-70001-40997