I lead software development at Weecology, University of Florida.

I build and run the open-source systems scientists use to install public ecological data, detect individual trees and birds in aerial imagery, and forecast how ecosystems change. I also maintain Using Databases and SQL, a core lesson for The Carpentries. I am a Google Summer of Code admin at NumFOCUS.

Weecology · GitHub · LinkedIn · Google Scholar · henrykironde@gmail.com

Ranked by how widely the work changes what people can measure, manage, and decide.

DeepForest is the Python package laboratories and agencies use to detect individual tree crowns and birds in airborne RGB imagery, then retrain the model on their own landscapes. It is the open implementation of that workflow, with a published tree model and a published bird model.

Documentation · Source · PyPI · conda-forge · Paper

MillionTrees is the benchmark that makes those models comparable. It holds more than a million tree annotations as boxes, polygons, and points, across sensors and forest types, so detection methods can be evaluated on the same global data.

Documentation · Source · PyPI

deepforestr is the R interface to DeepForest.

Wading birds are a measure of whether Everglades restoration is working. I maintain the pipeline that turns drone and aerial imagery into locations, species, and nests, and the long-term breeding record those forecasts depend on. The monitoring program has run for more than twenty years with Audubon Florida, Florida Atlantic University, and Everglades National Park.

Project site · Imagery workflow · Forecasts · Forecast site · Breeding data · Methods · Tools

The Data Retriever downloads, cleans, and installs public datasets into a database or flat files. A task that used to take hours or days becomes one command. It is a NumFOCUS affiliated project, with the same datasets available from Python, R, and Julia.

I am a Maintainer and Instructor for The Carpentries. I maintain Using Databases and SQL, the Software Carpentry lesson workshops use to teach researchers how to query data with SQL, built around an Antarctic survey dataset.

Lesson · Source · Maintainers · Instructors

The Portal Project has tracked rodents, ants, plants, and weather in the Chihuahuan Desert since 1977. I run the software that keeps that record usable and turns it into weekly forecasts of how the community will change.

Forecasts · Casting engine · Weekly runs · Forecast site source · Data · portalr on CRAN · Data explorer

I maintain the conda-forge packages researchers install in the field and on shared clusters: DeepForest, Retriever, slidingwindow, and keras-resnet. The lab site is weecology.org (source).

I am an admin and mentor for Google Summer of Code with NumFOCUS. Contributors from this program now work at Amazon, Bloomberg, Chicago Trading Company, Goldman Sachs, IBM, JPMorganChase, MathWorks, SigNoz, and research labs.

NumFOCUS GSoC · Data Retriever