BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//CERN//INDICO//EN
BEGIN:VEVENT
SUMMARY:High-Throughput Non-Destructive Prediction of Duckweed Protein Con
 tent Using Hyperspectral Imaging and Deep Learning
DTSTART;VALUE=DATE-TIME:20261001T080000Z
DTEND;VALUE=DATE-TIME:20261001T081500Z
DTSTAMP;VALUE=DATE-TIME:20260916T130456Z
UID:indico-contribution-2923@cern.ch
DESCRIPTION:Speakers: Weijuan Huang (Shenzhen University of Advanced Techn
 ology)\nDuckweed is a promising plant-based protein and biomass resource\,
  but efficient breeding and large-scale cultivation require rapid\, high-t
 hroughput\, and non-destructive evaluation of key traits. Conventional met
 hods for determining crude protein and chlorophyll contents are labor-inte
 nsive\, chemically demanding\, and destructive\, limiting their applicatio
 n in large-scale germplasm screening and real-time cultivation management.
  In this study\, we developed an integrated high-throughput phenotyping pl
 atform combining hyperspectral imaging (HSI) and deep learning for simulta
 neous non-destructive prediction of crude protein content\, leaf area-base
 d relative growth rate (RGR)\, and chlorophyll content in duckweed. More t
 han 200 duckweed samples cultivated in standardized 12-well microplates we
 re scanned using an HSI system covering 400–2000 nm with a spectral reso
 lution of 3 nm. A DeepLab v3+ semantic segmentation model was used to deli
 neate duckweed fronds and quantify leaf area\, enabling automated extracti
 on of growth-related phenotypes. A Transformer-based spectral-spatial feat
 ure fusion model was further developed to integrate RGB-derived spatial fe
 atures with hyperspectral signatures for biochemical and growth trait pred
 iction. The platform achieved strong predictive performance for crude prot
 ein content (R² = 0.938\, RMSE = 1.030\, prediction accuracy = 95.8%)\, c
 hlorophyll content (R² = 0.893\, RMSE = 1.514\, prediction accuracy = 94.
 7%)\, and leaf area-based RGR (prediction accuracy = 98.5%). With a throug
 hput of up to 120 samples per hour\, this HSI- and deep learning-based pla
 tform provides an efficient tool for duckweed germplasm evaluation\, high-
 protein strain selection\, and precision cultivation management.\n\nhttps:
 //indico.unina.it/event/117/contributions/2923/
LOCATION:Department of Agricultural Sciences of the University of Napoli F
 ederico II\, Portici\, Italy Sala Cinese
URL:https://indico.unina.it/event/117/contributions/2923/
END:VEVENT
END:VCALENDAR