Tools
BSTPP: Bayesian Spatiotemporal Point Process
BSTPP is a Python package for Bayesian inference on spatiotemporal point processes. It offers three different models: Log Gaussian Cox, Hawkes, and Cox Hawkes processes. The package includes a flexible pre-trained Variational Autoencoder (VAE) to accelerate posterior inference on Gaussian processes. Users can employ predefined trigger parameterizations or implement custom trigger functions using the extendable Trigger module.
https://www.georgemohler.com/_files/ugd/9226cc_a9bc736f02bb41fa92f7021aac6fc0be.pdf
EasyTPP
EasyTPP is a comprehensive benchmarking tool designed to advance research in temporal point processes (TPPs), which are crucial for modeling continuous-time event sequences across domains like healthcare, finance, and social networks. The motivation behind EasyTPP is to address the lack of a standardized framework, which impedes fair comparison, reproducibility, and progress in TPP research. The tool offers a unified interface for datasets, an array of evaluation methods, implementations of popular neural TPP models, and modular components for building and extending new models, supporting both PyTorch and TensorFlow frameworks. Some of the included datasets are synthetic Hawkes process data, Amazon user reviews, NYC taxi events, Taobao user clicks, Retweet sequences, and StackOverflow badges.
Forest Typology
Forest Typology is an AI-driven initiative by DeepMind, in partnership with the World Resources Institute and Google teams, aimed at mapping and classifying global forest types. The project goes beyond simple forest cover detection by estimating forest composition and identifying biodiversity hotspots and major carbon sinks. It provides forest land cover layers and benchmark datasets to advance research on climate change mitigation and conservation using geospatial AI.
GIS OPS
This tool provides routing information from sources like OpenStreetMap (OSM), TomTom, and HERE.
GeeFlow
GeeFlow is a DeepMind library for generating large-scale geospatial datasets using Google Earth Engine (GEE). It includes utilities, configuration templates, and pipeline scripts to streamline the creation of Earth observation datasets. While not designed for production deployment, GeeFlow is tailored to support geospatial AI research, particularly in conjunction with model training frameworks like Jeo.
.jpg)
Generative Neural Temporal Point Proces (GNTPP)
This tool introduces a PyTorch implementation of GNTPP in the paper GNTPP: `Exploring Generative Neural Temporal Point Process' (TMLR). This focuses on deep generative models for generating high-quality samples in the context of event occurrence modeling for TPP.
GeoPy
GeoPy is a library that provides an API; it is easy for developers to locate the coordinates of addresses, cities, countries, and landmarks across the globe using third-party geocoders and other data sources.
GeoStats.jl
GeoStats.jl is an extensible framework for geospatial data science and geostatistical modeling, fully written in Julia. It provides a unified interface for various geostatistical algorithms, including kriging, simulation, and estimation methods. The framework supports advanced geometric processing and is designed for high-performance geospatial computations.
Jeo
Jeo is a foundation model framework developed by DeepMind for embedding geospatial and temporal data into a shared representation space. It supports tasks like land cover classification and change detection by aligning diverse satellite imagery and related datasets. Jeo enables scalable, multimodal learning across different sensor types and geographies, facilitating more accurate and transferable models in geospatial AI.
PyTorch Geometric Temporal
This tool introduces various dynamic and temporal geometric deep learning, embedding, and spatio-temporal regression methods from a variety of published research papers.
https://github.com/benedekrozemberczki/pytorch_geometric_temporal
STPP Simulator
This GitHub project provides a framework for simulating spatio-temporal point processes, including Poisson and Hawkes models, on continuous spatial domains over time. It's useful for researchers who want to test inference methods or study synthetic event dynamics in controlled environments.
https://github.com/meowoodie/Spatio-Temporal-Point-Process-Simulator
STPPGC
STPPGC is flexible benchmarking toolkit for streaming Spatio-Temporal Point-Process (STPP) models. BenchSTPP is a modular, research-grade framework for end-to-end development, training, and evaluation of STPP models. It couples declarative YAML configuration with PyTorch Lightning execution, Ray Tune hyper-parameter optimisation, and version-controlled logging to deliver rapid prototyping and rigorous, reproducible benchmarking on streaming event data.
Spatial-Kfold
Spatial-Kfold is a Python library that improves cross-validation in spatial studies by offering spatial clustering and block resampling techniques. It facilitates "Leave Region Out" cross-validation, aiding in model generalization to new locations and improving the reliability of feature selection and hyperparameter tuning.
Spatstat
Spatstat is a comprehensive R package for analyzing spatial point patterns. It supports exploratory data analysis, model fitting (e.g., Poisson, Cox, Gibbs processes), spatial covariates, and simulation. The package handles irregular windows, marks, and inhomogeneous patterns, making it widely used in ecological, epidemiological, and environmental applications.
Tick Library
Tick is a Python library for statistical learning with a focus on time-dependent modeling, including Hawkes processes, generalized linear models, and survival analysis. It provides fast solvers and tools for point process simulation and inference, especially suited for financial, criminological, and event-based data.
mlr3spatiotempcv
mlr3spatiotempcv is an R package that extends the mlr3 machine learning framework with spatiotemporal resampling methods. It integrates various spatiotemporal partitioning strategies, facilitating model assessment, selection, and hyperparameter tuning for spatial and spatiotemporal data. The package provides a consistent interface for implementing state-of-the-art resampling techniques, aiding in the development of robust predictive models.