Deep learning models for crop yield prediction using remote sensing data, with CNN/LSTM and Gaussian Process approaches.
Crop Yield Prediction with Deep Learning is a research implementation for predicting agricultural yields using satellite imagery and machine learning. It provides tools for downloading remote sensing data from Google Earth Engine, preprocessing agricultural datasets, and training deep learning models including CNN/LSTM architectures and Gaussian Process regression. The project addresses food security challenges by enabling data-driven yield forecasting.
Researchers and data scientists working in agricultural technology, computational sustainability, and remote sensing applications. It's particularly relevant for those developing predictive models for crop monitoring and yield forecasting.
This implementation offers a complete pipeline from data acquisition to model training specifically designed for agricultural applications, with proven performance demonstrated through research awards. It provides efficient methods for handling large-scale satellite imagery and implements both deep learning and probabilistic modeling approaches.
Crop Yield Prediction with Deep Learning
Open-Awesome is built by the community, for the community. Submit a project, suggest an awesome list, or help improve the catalog on GitHub.
The project's method concatenates multi-year satellite images for bulk download from Google Earth Engine, speeding up data acquisition significantly, as described in the README.
It provides a full workflow from data preprocessing to model training and visualization, specifically designed for agricultural yield prediction with remote sensing data.
The underlying paper won a Best Student Paper award and a Big Data Innovation Challenge, indicating robust performance and relevance in computational sustainability.
Implements both CNN/LSTM deep learning and Gaussian Process regression, allowing for flexibility in handling temporal and spatial patterns with uncertainty quantification.
The code uses TensorFlow v0.9, which is obsolete and may cause compatibility issues with modern libraries or require significant updates for current projects.
The semi-supervised extension in the project is noted as not working well, as stated in the README, limiting its utility for advanced applications.
Requires manual data export from Google Drive and extensive preprocessing steps, making initial deployment cumbersome compared to automated pipelines.