Abstract
Abstract Accurate prediction of grain yield (GY) remains a major challenge in plant breeding due to complex interactions between genotype, environment, and management (G × E × M) factors. Remote sensing data from unmanned aerial vehicles (UAVs) equipped with multispectral sensors have emerged as a pivotal resource for high‐throughput phenotyping. In this study, we applied a deep transfer learning (DTL) approach to enhance GY prediction using UAV‐derived spectral and textural traits across multiple environments and developmental stages of winter wheat ( Triticum aestivum L.). The model's transferability and generalizability were evaluated across years, locations, nurseries, and stage‐specific scenarios. We employed fine‐tuning, which involves retraining a pretrained one‐dimensional convolutional neural network (1D‐CNN) model on scenario‐specific target data to enhance generalizability. Fine‐tuning was tested with 20%, 40%, 60%, and 80% of the target data to identify an optimal balance between model accuracy and adaptation, and compared with 1D‐CNN without DTL (baseline model). In cross‐year predictions, the baseline model (from 2022) performed poorly ( R 2 = −2.9 in 2023), while DTL improved prediction to an R 2 of 0.83 with 20% fine‐tuning, demonstrating strong temporal adaptability. Similarly, in cross‐location scenarios, baseline model performance was poor ( R 2 ranging from −15.3 to −0.4) but improved to 0.29–0.69 with 40% fine‐tuning. Stage‐specific predictions benefited most at Feekes stages 10.5 and 11, where DTL achieved an R 2 of 0.78 and 0.82, respectively, compared to baseline model R 2 values near 0. These results demonstrate that DTL improves model transferability and generalizability, increasing prediction accuracy and offering a resource‐efficient tool to accelerate selection for complex and labor‐intensive traits in modern breeding programs.
Abstract from the paper’s OpenAlex record; open copy licensed cc-by. Read at the source ↗
The text the model read, with the source record’s inline markup (italics, sub- and superscripts) rendered and its character entities decoded; quote offsets refer to the raw text.
Claims 7
Claim · model-written summary
Cross-year baseline prediction had R² −2.9 in 2023.
- Taxon
- winter wheat
Triticum aestivumt0 - Outcome
- baseline modele0
- Value
- R 2 = −2.9e0
- Qualifier
- in 2023e0
- Comparator
- baseline modele0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“the baseline model (from 2022) performed poorly ( R 2 = −2.9 in 2023)”
t0“winter wheat across temporal and spatial variability”
Claim · model-written summary
With 20% fine tuning, cross-year prediction improved to R² 0.83.
- Taxon
- winter wheatTriticum aestivumt0
- Treatment
- 20% fine‐tuninge0
- Outcome
- predictione0
- Direction
- improvede0
- Value
- R 2 of 0.83e0
Not reportedComparator
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“DTL improved prediction to an R 2 of 0.83 with 20% fine‐tuning”
t0“winter wheat across temporal and spatial variability”
Claim · model-written summary
Cross-location baseline R² ranged from −15.3 to −0.4.
- Taxon
- winter wheatTriticum aestivumt0
- Outcome
- baseline model performancee0
- Value
- R 2 ranging from −15.3 to −0.4e0
- Comparator
- baseline modele0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“baseline model performance was poor ( R 2 ranging from −15.3 to −0.4)”
t0“winter wheat across temporal and spatial variability”
Claim · model-written summary
With 40% fine tuning, cross-location R² improved to 0.29–0.69.
- Taxon
- winter wheatTriticum aestivumt0
- Treatment
- 40% fine‐tuninge0
- Outcome
- model performancee0
- Direction
- improvede0
- Value
- 0.29–0.69e0
- Comparator
- baseline modele0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“Similarly, in cross‐location scenarios, baseline model performance was poor ( R 2 ranging from −15.3 to −0.4) but improved to 0.29–0.69 with 40% fine‐tuning”
t0“winter wheat across temporal and spatial variability”
Claim · model-written summary
At Feekes 10.5, transfer learning reached R² 0.78 versus baseline near zero.
- Taxon
- winter wheatTriticum aestivumt0
- Organ / stage
- Feekes stages 10.5e0
- Outcome
- R 2e0
- Value
- 0.78e0
- Qualifier
- near 0e0
- Comparator
- baseline modele0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“at Feekes stages 10.5 and 11, where DTL achieved an R 2 of 0.78 and 0.82, respectively, compared to baseline model R 2 values near 0”
t0“winter wheat across temporal and spatial variability”
Claim · model-written summary
At Feekes 11, transfer learning reached R² 0.82 versus baseline near zero.
- Taxon
- winter wheatTriticum aestivumt0
- Organ / stage
- 11e0
- Outcome
- R 2e0
- Value
- 0.82e0
- Qualifier
- near 0e0
- Comparator
- baseline modele0
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“at Feekes stages 10.5 and 11, where DTL achieved an R 2 of 0.78 and 0.82, respectively, compared to baseline model R 2 values near 0”
t0“winter wheat across temporal and spatial variability”
Claim · model-written summary
The authors conclude transfer learning improves accuracy and offers a resource efficient breeding tool.
- Taxon
- winter wheatTriticum aestivumt0
- Outcome
- prediction accuracye0
- Direction
- increasinge0
Not reportedComparator
Organisms pictured · Representative pictures of the matched species, not figures from the paper.
Verbatim
e0“DTL improves model transferability and generalizability, increasing prediction accuracy and offering a resource‐efficient tool to accelerate selection for complex and labor‐intensive traits in modern breeding programs”
t0“winter wheat across temporal and spatial variability”