# RESEARCH — publications. One block per paper.
#
#   logo:  the publisher mark on the left. Format: path or URL = alt text.
#          Marks want a transparent background; the disc supplies the fill.
#   doi:   the paper's DOI, e.g. `doi: 10.1038/s41598-025-00000-0`. This one
#          line prints the DOI on the card AND builds the "Read the paper"
#          button, so the two can never disagree. Until a real DOI is here,
#          the card shows no DOI and no button, because a dead button reads
#          worse than none. Paste it bare, as `doi:10.…`, or as a full
#          https://doi.org/… URL; all three are accepted.
#   link:  a button of your own. Format: label -> url. A block that has one
#          uses it instead of the DOI button, for a paper whose DOI is printed
#          on the proofs but not live at doi.org yet. Delete the line the day
#          it resolves and the DOI button comes back by itself.
#
# The "All publications on Google Scholar" link at the bottom comes from
# `scholar:` in site.md and only appears once that line has a URL.
template: research
label: RESEARCH
title: Publications

Two journal papers and one conference paper, all on applied vision and sequence models.

---
title: A Modular Vision Transformer and Machine Learning Framework for Scalable Multi-Crop Plant Disease Classification
meta: Scientific Reports · Nature Portfolio · Aug 2026
logo: assets/publications/nature.png = Nature Portfolio
doi: 10.1038/s41598-026-65004-7

A modular framework pairing a vision transformer backbone with classical machine learning heads, built to classify disease across multiple crops rather than one at a time.

---
title: Hybrid Deep Temporal Network for Animal Activity Recognition with Self-Attention and Folded Recurrent Learning
meta: Applied Computational Intelligence and Soft Computing · Wiley · Aug 2026 (In production)
logo: assets/publications/wiley.png = Wiley
doi: 10.1155/acis/2028283
link: Read the paper -> https://authors.wiley.com/index.html/article/100655703

Sequence model for activity recognition from accelerometer data, combining folded recurrent learning with self-attention over sliding time windows.

---
title: Smart Retail Billing System using YOLOv10 Object Detection and Raspberry Pi
meta: ICOECA 2025 · IEEE Xplore · March 2025
logo: https://cdn.simpleicons.org/ieee/00629B = IEEE
doi: 10.1109/ICOECA66273.2025.00186

Real-time edge product detection for retail checkout, presented at the International Conference on Expert Clouds and Applications.