Crop monitoring using remote sensing land use and land change data: Comparative analysis of deep learning methods using pre-trained CNN models

Big Data Research, Journal Year: 2024, Volume and Issue: 36, P. 100448 - 100448

Published: March 20, 2024

Language: Английский

---

Deep Transfer Learning Using Real-World Image Features for Medical Image Classification, with a Case Study on Pneumonia X-ray Images

Bioengineering, Journal Year: 2024, Volume and Issue: 11(4), P. 406 - 406

Published: April 20, 2024

Deep learning has profoundly influenced various domains, particularly medical image analysis. Traditional transfer approaches in this field rely on models pretrained domain-specific datasets, which limits their generalizability and accessibility. In study, we propose a novel framework called real-world feature learning, utilizes backbone initially trained large-scale general-purpose datasets such as ImageNet. We evaluate the effectiveness robustness of approach compared to from scratch, focusing task classifying pneumonia X-ray images. Our experiments, included converting grayscale images RGB format, demonstrate that real-world-feature consistently outperforms conventional training across performance metrics. This advancement potential accelerate deep applications imaging by leveraging rich representations learned models. The proposed methodology overcomes limitations models, thereby enabling accelerated innovation diagnostics healthcare. From mathematical perspective, formalize concept provide rigorous formulation problem. experimental results empirical evidence supporting approach, laying foundation for further theoretical analysis exploration. work contributes broader understanding transferability domains significant implications development accurate efficient analysis, even resource-constrained settings.

Language: Английский

---

Predicting spatial and temporal variability in soybean yield using deep learning and open source data

European Journal of Agronomy, Journal Year: 2025, Volume and Issue: 164, P. 127498 - 127498

Published: Jan. 15, 2025

Language: Английский

---

Crossformer-Based Model for Predicting and Interpreting Crop Yield Variations Under Diverse Climatic and Agricultural Conditions

Agriculture, Journal Year: 2025, Volume and Issue: 15(9), P. 958 - 958

Published: April 28, 2025

Crop yield prediction is critical for agricultural decision making and food security. Traditional models struggle to capture the complex interactions among meteorological, soil, factors. This study introduces Crossformer, a Transformer-based model with Local Perception Unit (LPU) local dependencies Cross-Window Attention Mechanism global dependencies. Experiments on winter wheat, rice, corn datasets show that Crossformer outperforms CNN, LSTM, Transformer in Test Loss, R2, MSE, MAE. For instance, dataset, achieves Loss of 0.0271 an R2 0.9863, compared 0.7999 0.1634 respectively, demonstrating substantial improvement predictive performance. Interpretability analysis highlights importance temperature precipitation prediction, aligning insights. The results demonstrate Crossformer’s potential precision agriculture.

Language: Английский

---

TCNT: a temporal convolutional network-transformer framework for advanced crop yield prediction

Journal of Applied Remote Sensing, Journal Year: 2024, Volume and Issue: 18(04)

Published: Nov. 5, 2024

Modern agricultural practices require accurate prediction of crop yields, particularly in the face a changing climate. Despite improvement estimating yield over years through machine learning (ML) algorithms, increased volatility and complexity weather patterns continue to make use conventional ML models unreliable for understanding intricate relationships. We therefore explore potential advanced ML, specifically transformer-based architecture coupled with temporal convolutional network (TCN) prediction. argue that transformers' ability model long-range dependencies within data sequences makes them well suited handle complex relationships comprehensive datasets. In addition, integration TCN would complement transformer's strengths by focusing on feature extraction. Alongside climatic variables, study integrates soil properties, moderate resolution imaging spectroradiometer (MODIS), average analyze factors influencing growth yield. The proposed TCN-transformer (TCNT) is trained evaluated using corn soybean values selected from 264 counties Illinois, Iowa, Wisconsin (the United States belt region). Furthermore, experimental results show superior performance our TCNT framework other state-of-the-art both in-season end-of-season predictions.

Language: Английский

---

MHRA-MS-3D-ResNet-BiLSTM: A Multi-Head-Residual Attention-Based Multi-Stream Deep Learning Model for Soybean Yield Prediction in the U.S. Using Multi-Source Remote Sensing Data

Remote Sensing, Journal Year: 2024, Volume and Issue: 17(1), P. 107 - 107

Published: Dec. 31, 2024

Accurate prediction of soybean yield is important for safeguarding food security and improving agricultural management. Recent advances have highlighted the effectiveness ability Machine Learning (ML) models in analyzing Remote Sensing (RS) data this purpose. However, most these do not fully consider multi-source RS prediction, as processing increases complexity limits their accuracy generalizability. In study, we propose Multi-Residual Attention-Based Multi-Stream 3D-ResNet-BiLSTM (MHRA-MS-3D-ResNet-BiLSTM) model, designed to integrate various types, including Sentinel-1/2 imagery, Daymet climate data, soil grid information, improved county-level U.S. prediction. Our model employs a multi-stream architecture process diverse types concurrently, capturing complex spatio-temporal features effectively. The 3D-ResNet component utilizes 3D convolutions residual connections pattern recognition, complemented by Bidirectional Long Short-Term Memory (BiLSTM) enhanced long-term dependency learning arrangements forward backward directions. An attention mechanism further refines model’s focus dynamically weighting significance different input efficient We trained MHRA-MS-3D-ResNet-BiLSTM using datasets from 2019 2020 evaluated its performance with 2021 2022. results demonstrated robustness adaptability unseen achieving an R2 0.82 Mean Absolute Percentage Error (MAPE) 9% 2021, 0.72 MAPE 12% This surpassed some state-of-the-art like MS-3D-ResNet-BiLSTM, other traditional ML methods Random Forest (RF), XGBoost, LightGBM. These findings highlight methodology’s capability handle multiple role predictions, which can be helpful sustainable agriculture.

Language: Английский

---

Changes in the leaf area-seed yield relationship in soybean driven by genetic, management and environments: Implications for High-Throughput Phenotyping

in silico Plants, Journal Year: 2024, Volume and Issue: 6(2)

Published: Jan. 1, 2024

Abstract High-throughput crop phenotyping (HTP) in soybean (Glycine max) has been used to estimate seed yield with varying degrees of accuracy. Research this area typically makes use different machine-learning approaches predict based on images a strong focus analytics. On the other hand, significant part breeding community still utilizes linear relate canopy traits and relying parsimony. Our research attempted address limitations related interpretability, scope system comprehension inherent previous modelling approaches. We utilized combination empirical simulated data augment experimental footprint as well explore combined effects genetics (G), environments (E) management (M). flexible functions without assuming pre-determined response between yield. Factors such maturity date, duration vegetative reproductive periods, harvest index, potential leaf size, planting date plant population affected shape canopy-seed relationship optimum values at which selection high yielding genotypes should be conducted. This work demonstrates that there are avenues for improved application HTP programs if similar considered.

Language: Английский

---

Fusion of MobileNet and GRU: Enhancing Remote Sensing Applications for Sustainable Agriculture and Food Security

Remote Sensing in Earth Systems Sciences, Journal Year: 2024, Volume and Issue: unknown

Published: Dec. 19, 2024

Language: Английский

---

From Tradition to Transformation: Deep and Self-Supervised Learning Approaches for Remote Sensing in Agriculture and Environmental Change

Published: Jan. 1, 2024

Language: Английский

---